diversity of thought in the blogosphere: …diversity of thought in the blogosphere: implications...
Post on 21-May-2020
3 Views
Preview:
TRANSCRIPT
DIVERSITY OF THOUGHT IN THE BLOGOSPHERE: IMPLICATIONS FOR
INFLUENCING AND MONITORING IMAGE
A Dissertation
by
PAUL DWYER
Submitted to the Office of Graduate Studies of
Texas A&M University
in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
August 2008
Major Subject: Marketing
DIVERSITY OF THOUGHT IN THE BLOGOSPHERE: IMPLICATIONS FOR
INFLUENCING AND MONITORING IMAGE
A Dissertation
by
PAUL DWYER
Submitted to the Office of Graduate Studies of
Texas A&M University
in partial fulfillment of the requirements for the degree of
DOCTOR OF PHILOSOPHY
Approved by:
Chair of Committee, Rajan Varadarajan
Committee Members, Benton Cocanougher
Venkatesh Shankar
Richard Woodman
Head of Department, Jeffrey Conant
August 2008
Major Subject: Marketing
iii
ABSTRACT
Diversity of Thought in the Blogosphere: Implications for Influencing and Monitoring
Image.
(August 2008)
Paul Dwyer, B.S.C.E., University of South Florida;
M.B.A., Texas A&M University
Chair of Advisory Committee: Dr. Rajan Varadarajan
A blog, a shortened form of weblog, is a website where an author shares
thoughts in posts or entries. Most blogs permit readers to add comments to posts and
thereby be a conversational mechanism. One way that companies have started to use
blogs is to monitor their corporate image (in this dissertation, the term image is used in
reference to corporate, brand and/or product image). This study focuses on how common
socio-psychological processes mediate consumers’ revelation of corporate image in the
blogosphere. Centering resonance analysis, a means of measuring similarity between
two bodies of text, is used in conjunction with multidimensional scaling to locate text as
cognitive objects in a space. Clusters are then detected and measured to quantify
diversity in the thoughts expressed. Detected patterns are studied from a social process
theory perspective, where complex phenomena are hypothesized to be the result of the
interaction of simpler processes.
iv
A majority of blog commenters compromise the expression of their thoughts to
gain social acceptance. This study identifies the most extreme of such people so
companies who monitor blogs can assign less weight to image indications gained from
them as they may be merely expressing thoughts that are intended to maintain social
acceptance.
It was also found that single-theme blogs attract a readership with similarly
narrow interests. The boldest and most diverse thinkers among comment writers have the
most impact because of their ability to provoke the thinking of others. However,
commenters who repeat the same ideas have little effect, suggesting that introducing
shills is unlikely to shift the sentiment of a blog’s readership.
People participate in blog communities for reasons (e.g., need for community)
that may undermine thought diversity. However, there may be value in serving those
needs even though no valuable insights are provided into image or directions for product
development. Members of homogeneous-thinking communities were observed to more
actively participate, with greater longevity. This may increase loyalty to the company
hosting the blog.
v
ACKNOWLEDGEMENTS
I would like to thank my committee chair, Dr. Varadarajan, and my committee
members, Dr. Cocanougher, Dr. Shankar, and Dr. Woodman, for their guidance and
support throughout the course of this research.
Thanks also go to my friends, colleagues and the department faculty and staff for
making my time at Texas A&M University a great experience.
vi
TABLE OF CONTENTS
Page
ABSTRACT .............................................................................................................. iii
ACKNOWLEDGEMENTS ...................................................................................... v
TABLE OF CONTENTS .......................................................................................... vi
LIST OF FIGURES................................................................................................... viii
LIST OF TABLES .................................................................................................... xi
CHAPTER
I INTRODUCTION………………………............................................ 1
Literature Overview ....................................................................... 7
Research Questions ........................................................................ 9
Potential Contributions................................................................... 11
II OVERVIEW OF RELEVANT LITERATURE................................... 12
Cognitive Associations................................................................... 12
Image Management ........................................................................ 18
Detecting and Extracting Cognitive Associations.......................... 20
Mapping Cognitive Associations ................................................... 25
Social Process Theory .................................................................... 30
Summary ........................................................................................ 44
III CONCEPTUAL MODEL AND HYPOTHESES ................................ 45
Cognitive Diversity ........................................................................ 45
Cultural Tribalism, Need-for-Cognition and Cognitive Diversity . 47
Threshold Models of Collective Behavior and
Cognitive Diversity ........................................................................ 52
Flocking Theory and Cognitive Diversity...................................... 53
Memetics and Cognitive Diversity................................................. 57
Reciprocity and Cognitive Diversity.............................................. 66
Alternative Model .......................................................................... 71
Relevance of Sociological Theory ................................................. 75
vii
CHAPTER Page
Summary ........................................................................................ 77
IV METHODOLOGY............................................................................... 82
Data ................................................................................................ 83
Method ........................................................................................... 87
Limitations ..................................................................................... 115
Summary ........................................................................................ 116
V RESULTS............................................................................................. 117
Classification of Blog Entry Comments......................................... 118
Classification of Commenters ........................................................ 125
Summary ........................................................................................ 144
VI DISCUSSION AND CONCLUSIONS................................................ 145
Review of Research Questions....................................................... 145
Implications for Research in Marketing......................................... 160
Managerial Implications................................................................. 167
Public Policy Implications ............................................................. 177
Limitations ..................................................................................... 177
Summary ........................................................................................ 179
REFERENCES.......................................................................................................... 181
APPENDIX A ........................................................................................................... 193
APPENDIX B ........................................................................................................... 215
APPENDIX C ........................................................................................................... 219
APPENDIX D ........................................................................................................... 230
VITA ......................................................................................................................... 238
viii
LIST OF FIGURES
FIGURE Page
1.1 Corporate Image Management and the Blogosphere .............................. 7
1.2 Conceptual Map of the Literature Reviewed .......................................... 9
2.1 Corporate Image Management with Blogs.............................................. 19
2.2 Betweenness Centrality in Text............................................................... 24
2.3 The Power-Law Distribution of Attention in the Blogosphere ............... 33
3.1 Cognitive Diversity Latent Variable ....................................................... 46
3.2 Cultural Tribalism and Need-for-Cognition as Opposing Motivations .. 48
3.3 Cultural Tribalism / Need-for-Cognition Reflective Sub-Models .......... 51
3.4 Flocking and Idea Seeding Reflective Sub-Models ................................ 56
3.5 Spectrum of Blog Author Thought Diversity.......................................... 58
3.6 Locus of Influence Conceptual Map ....................................................... 60
3.7 Evangelists and Idea Seeders .................................................................. 61
3.8 Perceptual Map of Commenter Types..................................................... 62
3.9 Memetic Propagation Reflective Sub-Models ........................................ 65
3.10 Naïve Model: Hypothesized Relationships ............................................. 71
3.11 Alternative Model: Relationships............................................................ 73
4.1 Pretest Blog Entry Comment Mapping ................................................... 92
4.2 Pretest BFA Commenter Clusters ........................................................... 94
4.3 Pretest GM Fastlane Commenter Clusters .............................................. 95
ix
FIGURE Page
4.4 Pretest Normal Probability Plot of BFA Cognitive Diversity................. 103
4.5 Pretest Normal Probability Plot of GM Fastlane Cognitive Diversity.... 104
4.6 Naïve Factor Model Fitted with Pretest Data.......................................... 108
4.7 Alternative Factor Model Fitted with Pretest Data ................................. 110
4.8 PC Search DAG ...................................................................................... 113
4.9 Greedy Equivalent Search DAG ............................................................. 114
4.10 PC Search and GES Commonality DAG ................................................ 115
5.1 Blog Entry Comment Mapping ............................................................... 119
5.2 Blog Cultural Tribalism Histogram......................................................... 120
5.3 Paneled Histograms of Thought Diversity .............................................. 122
5.4 Paneled Histograms of Cultural Tribalism and Need for Cognition....... 123
5.5 All Blog Commenter Clusters ................................................................. 126
5.6 Paneled Histograms of Boldness and Individual Thought Diversity ...... 127
5.7 All Blog Idea Seeding Histogram ........................................................... 128
5.8 Individual Blog Commenter Clusters...................................................... 129
5.9 Paneled Histograms of Blog Author and Commenter Types .................. 131
5.10 Fitted Core Naïve Model......................................................................... 139
5.11 Fitted Alternative Model ......................................................................... 141
x
FIGURE Page
A-1 Dendrogram of an Agglomerative Hierarchical Cluster ......................... 193
A-2 The Long Tail of the Blogosphere .......................................................... 204
A-3 A Social Attractor in Individual Thought Diversity................................ 212
C-1 Individual Blog Normal Probability Plots of the
Cognitive Diversity Factor ...................................................................... 226
C-2 Full Reflective and Formative Naïve Model........................................... 228
C-3 Alternative Model of Potential Interrelationships ................................... 239
D-1 Naïve Model Cognitive Diversity ........................................................... 230
D-2 Naïve Model Indoctrination / Free Thought............................................ 230
D-3 Naïve Model Cultural Tribalism ............................................................. 231
D-4 Naïve Model Idea Seeding ...................................................................... 231
D-5 Naïve Model Flocking............................................................................. 232
D-6 Naïve Model Mass Dissemination .......................................................... 232
D-7 Naïve Model Reciprocation .................................................................... 233
D-8 Naïve Model Need for Cognition............................................................ 233
D-9 Alternative Model Cognitive Diversity................................................... 234
D-10 Alternative Model Indoctrination / Free Thought ................................... 234
D-11 Alternative Model Cultural Tribalism..................................................... 235
D-12 Alternative Model Idea Seeding.............................................................. 235
D-13 Alternative Model Flocking .................................................................... 236
D-14 Alternative Model Mass Dissemination.................................................. 236
D-15 Alternative Model Need for Cognition ................................................... 237
xi
LIST OF TABLES
TABLE Page
1.1 A Hierarchy of Marketing-Relevant Blog Types .................................... 4
3.1 Hypotheses and Construct Definition Summary..................................... 79
4.1 Data Sources............................................................................................ 85
4.2 Data Source Quantities............................................................................ 86
4.3 Pretest Confirmatory Factor Analysis, Reliability, Variance Explained
and Factorial Invariance .......................................................................... 99
4.4 Pretest Correlation and Φ-matrix of Primary Constructs ........................ 101
4.5 Pretest Multigroup Fitted Naïve Model Parameter Values ..................... 106
4.6 Pretest Multigroup Fitted Alternative Model Parameter Values............. 107
4.7 Pretest Alternative Model Tests of Mediation ........................................ 109
5.1 Comment Discriminant Function Coefficients (Equation 3.1) ............... 124
5.2 Commenter Discriminant Function Coefficients (Equation 3.2) ............ 132
5.3 Normalized Estimates of Mardia’s (1970) Multivariate Coefficient ...... 134
5.4 Confirmatory Factor Analysis, Reliability, Variance Explained and
Factorial Invariance................................................................................. 136
5.5 Multigroup Fitted Naïve Model Parameter Values ................................. 140
5.6 Multigroup Fitted Alternative Model Parameter Values......................... 140
5.7 Alternative Model: Tests of Mediation ................................................... 142
5.8 Summary of Hypotheses and Results...................................................... 143
C-1 Correlation and Φ-matrix of Primary Constructs for AutoBlog ............. 219
xii
TABLE Page
C-2 Correlation and Φ-matrix of Primary Constructs for Blog for America . 219
C-3 Correlation and Φ-matrix of Primary Constructs for Blog Maverick ..... 220
C-4 Correlation and Φ-matrix of Primary Constructs for The Consumerist.. 220
C-5 Correlation and Φ-matrix of Primary Constructs for EnGadget ............. 221
C-6 Correlation and Φ-matrix of Primary Constructs for
The Evangelical Outpost ......................................................................... 221
C-7 Correlation and Φ-matrix of Primary Constructs for GM Fastlane ........ 222
C-8 Correlation and Φ-matrix of Primary Constructs for Freakonomics....... 222
C-9 Correlation and Φ-matrix of Primary Constructs for Gizmodo .............. 223
C-10 Correlation and Φ-matrix of Primary Constructs for
Google Blogoscoped ............................................................................... 223
C-11 Correlation and Φ-matrix of Primary Constructs for Joystiq .................. 224
C-12 Correlation and Φ-matrix of Primary Constructs for PaulStamatiou...... 224
C-13 Correlation and Φ-matrix of Primary Constructs for Townhall .............. 225
C-14 Correlation and Φ-matrix of Primary Constructs for TV Squad............. 225
C-15 Correlation and Φ-matrix of Primary Constructs for
The Unofficial Apple Weblog................................................................. 226
1
CHAPTER I
INTRODUCTION
The term blog1, a shortened form of weblog, refers to a website where an author
shares thoughts in posts or entries. Most blogs permit readers to add comments to posts
and thereby be a conversational mechanism. Popular blogs have become meeting places
for communities of readers. Organizations are trying to better understand the relevance
of these online communities to their operations and how they should respond to the
phenomenon.
Murray (2007) ranked blog readership as a percentage of population among
world nations. He reported that Japan ranked highest with 74% of the population reading
blogs, followed by South Korea (43%), China (39%), the United States (27%), and the
United Kingdom (23%) in the top five. In each country, the younger the reader, the more
time they spend reading blogs. Additionally, blog readers were more likely to self-
identify as taking part in political, community or social welfare activities and were thus
designated “influencers” (p. 5).
____________
This dissertation follows the style of Journal of Marketing. 1 The first use of a term central to this study is italicized and the word is defined in the
glossary in Appendix A.
2
Technorati (2007), a blog search engine, reports that it tracks almost 100 million
blogs. These blogs cover such a variety of subject matter that they resist the creation of
the kind of taxonomy that scientists generally seek to create. The blogosphere eschews
top-down taxonomies in favor of folksonomies, a form of collaborative categorization
based on the aggregation of tags, key words descriptive of content (Golder and
Huberman 2006). Nevertheless, in Table 1.1 a categorization framework is described
that delineates types of blogs suggested most of interest to the marketing community.
The framework begins by partitioning marketing-related blogs into those related
to business, and those related to public policy and non-profit issues, so as to capture the
full breadth of marketing areas of practice. While, it may seem that more importance has
been conferred on business blogs, as the category breakdown is more nuanced, the
general structure of the business blog hierarchy can be applied to public policy and non-
profit blogs as well.
Established information providers have been augmenting their current service
offerings with new ones based on the extraction of information from the blogosphere,
the collection of all blogs. One example is VNU, the corporate parent of ACNielsen and
Nielsen Media Research, which announced in January 2006 the acquisition of
BuzzMetrics and Intelliseek, two pioneers in the monitoring of online consumer-
generated media, and merged them into Nielsen BuzzMetrics. This acquisition of new
competencies by an industry giant signals that they perceive consumer-generated media
to be a new information resource that, rather than being a passing fad, is growing in
significance.
3
One type of information of considerable interest to companies is information
about their corporate image2. Companies have an image they would like to instill in the
minds of customers. They are aware that all methods of communicating that image are
imperfect and inefficient to varying degrees. Hence, the need to infer the image in the
consumers’ minds, compare it to the one they wish to create and then develop and
execute ways to bring the two images into congruence. Nielsen BuzzMetrics offers a
service called BrandPulse that monitors blogs and message boards for mention of their
client customers’ brands. Blogs of many types are potential sources for marketing
insights even though organizations tend to directly connect with consumers in only a few
blog types. As most of the world economies are either consumerist or in the process of
becoming so, any blog can be a venue for the discussion of corporate brands, trademarks
and other image assets.
2 Throughout this dissertation, the term image refers generally to corporate, brand and/or
product image.
4
TABLE 1.1
A Hierarchy of Marketing-Relevant Blog Types
Example and Description External Sanctioned Employee
IBM developerWorks. A large set of blogs authored by IBM
product developers from a wide array of product lines.
Senior Executive GM Fastlane. GM Vice Chairman Bob Lutz’ blog on current and
future product offerings.
Character The Family Guy Blog. Supposedly written by a fictional
character, often a TV character, mascot or brand icon.
Corporate
Internal A blog intended for the exchange, accumulation and management
of internal corporate knowledge.
Communal Hiptop Nation. A consumer’s blog that became the primary
“moblog” (i.e., mobile weblog) for Danger Hiptop and T-Mobile
Sidekick enthusiasts to post photos, etc.
Diary-like Wil Wheaton. An actor’s personal blog.
Firm-Sponsored Bob Balfe. An IBM Lotus developer’s blog, linked to by IBM,
but hosted apart from developerWorks.
Review-like Paul Stamatiou. The widely linked to blog of a college student
who reviews products and gives technical support.
Personal
Unsanctioned Employee
The Masked Blogger. Rumored to be the blog of an Apple
employee. Apple authorizes no external blogs. Corporate Watchdog WalMart Watch. Analysis and commentary of WalMart.
Consumer Advocacy The Consumerist. A consumer news and retaliation blog.
Business-Oriented
Third-Party
Industry-level AutoBlog. News and reviews of cars in general.
Religion and Culture The Evangelical Outpost. Commentary on culture and religion
from an evangelical perspective.
Social and Health Kalyn’s Kitchen. Recipes for healthy meals.
Public Policy and Non-Profit
Organizational and Personal
Political Blog for America. The official blog of the Democratic Party.
5
Scoble and Israel (2006), the authors of Naked Conversations, based on their
informal observations, note that blogs can be a rich source of insight into how customers
view a company. However, they caution that these insights may only reflect the views of
a vocal minority, a phenomenon they call the echo chamber. This expression of caution
inspired the general research question that started this study’s line of research: Is there a
way to measure the extent to which a blog has become an echo chamber?
The echo chamber phenomenon is one of many models or social forms that have
been used to explain behavior observed within group or community settings. The term
“social forms” is drawn from social process theory and refers specifically to lesser
socio-psychological processes that operate both in tandem and conjunction to create
complex social phenomena (Cederman 2005). This study investigates the expression of
thought in a diverse selection of corporate, political and individual weblogs from a social
process theory perspective.
Kawasaki (2004), an entrepreneur and venture capitalist, opines that the most
successful products are polarizing: consumers either love them or hate them; only
mediocre products are viewed with indifference. Kawasaki’s point of view is consistent
with Muniz and O’Guinn (2001) and McAlexander et al. (2002) who see passion as what
motivates community and results in loyalty. It might be thought then that the echo
chamber, and its attendant limited expression of thought, is the most desirable condition
for consumer communities as it allows companies to monitor their image among its core
constituency without the introduction of noise from those who passionately dislike, or
are merely indifferent to, them and their products. This argument has surface merit, but
6
since an echo chamber is an extreme situation of limited thought expression, it creates a
context that by focusing on the status quo engenders few ideas for new products or to
solve problems that might expand the customer base without alienating the core. These
limited perspectives also provide no indication of how a target market may be evolving
in its determinants of satisfaction.
Image monitoring is complementary to satisfaction analysis frameworks like
Kano’s (1984) model, where product attributes are differentiated on the basis of being
basic essentials, determinants of performance and sources of excitement. By maximizing
diversity of perspectives companies increase the likelihood of finding product
enhancements that are basic essentials to new consumer segments while being at worst,
factors of indifference, and at best, performance or excitement increasers to existing
customers. Additionally, it is naïve to believe that other companies will not compete for
their competitors’ loyal and most profitable core. Jones and Sasser (1995) note that for a
consumer to be truly loyal they must be completely satisfied. Furthermore, they note that
complete satisfaction is a moving target and that communing only with existing
customers can miss leading indicators of market shifts and their attendant satisfaction
determinants. They recommend that companies use a variety of methods to listen to
existing, potential and former customers. Clearly, the goal of complete satisfaction can
only be met by maximizing diversity of insight from the market.
Figure 1.1 shows how the blogosphere could fit into a corporate image
management process. Kapferer (2004) notes that corporate image can be affected by
conventional marketing communications such as public relations, advertising,
7
sponsorship and personal selling, as a part of a multifaceted image management
program. This study investigates the efficacy of adding blogging, as a new and emerging
form of interactive marketing communication, and blog monitoring to the array of image
management activities. Companies can monitor how they are portrayed in the
blogosphere and use those insights to modify their image building activities thus creating
a feedback loop. Consumers are assumed to interact with the blogosphere in a manner
mediated by social forms. This study broadly focuses on this mediation and its impact on
the extent to which the image in the blogosphere presents an accurate representation of
brand image in consumers’ minds.
FIGURE 1.1
Corporate Image Management and the Blogosphere
Literature Overview
This dissertation draws on an eclectic body of past research. Figure 1.2 provides
an overview of the literature underpinnings of this research. In Chapter II, each entity in
8
Figure 1.2 is described in the order given and greater detail is provided into how these
entities connect together. A more detailed definition of a blog is provided. In addition,
the large body of prior research into the nature of brand cognitive associations and how
these associations have previously been elicited from consumers in a context similar to
that of a blog is summarized. Since blogging uses written communications, advances in
textual analysis are also summarized, culminating in a detailed description of centering
resonance analysis (CRA) which yields a similarity measure for pairs of text bodies.
Then, how this similarity measure can be translated with multidimensional scaling into
points in a cognitive space and used as a basis for measuring diversity in the thoughts
expressed is described. Finally, some alternative models (social process theories) are
discussed that explain the diversity of thought expression, and thought expression in
general, in a group context.
9
FIGURE 1.2
Conceptual Map of the Literature Reviewed
Research Questions
As previously stated, companies are interested in knowing how they are
perceived by consumers and are turning to online community monitoring to gain
insights. However, the question of the extent to which perceptions gained from these
communities are a true reflection of their image in the overall market is not known and
has not been researched. Blog-based research can be characterized as quasi-experimental
because it has some of the attributes of experimental scientific research, like the
collection of data to which statistical methods can be applied, but it suffers from a lack
of control over factors that might confound the identification of testable cause-and-effect
relationships. Without an understanding of the processes that are operant within blogs,
10
not much faith can be placed in findings based on statistical analyses. This research
study is aimed at identifying some of the major processes operating within blogs,
particularly the ones that affect how thoughts get expressed and thereby perform some
basic research for the modeling of cause-and-effect within blogs. Toward this end, the
following research questions are investigated:
1. What mechanisms explain the expression of thoughts by individuals in a blog?
Since comments are a vehicle for consumers to express their thoughts, there is a
need to determine which social processes are most relevant in determining
whether or not comments are made, and if so, when they are made.
2. What processes expand cognitive diversity, the expression of diverse thought?
Following Scoble and Israel’s (2006) focus on the echo chamber phenomenon as
a factor that could reduce the expression of diversity in thought, research
questions 2 and 3 seek to look at both sides of the issue of factors affecting the
expression of thought diversity. Research question 2 focuses on which of the
processes identified in the first research question tend to increase the diversity of
thought expression.
3. What processes limit cognitive diversity? Addressing this question has the
potential to shed insights into Scoble and Israel’s (2006) concerns about the echo
chamber phenomenon and reveal other factors that might limit cognitive
diversity.
4. What is the relative importance of these processes? The core concept of social
process theory is that social processes act in tandem and in conjunction. It may
11
be easier to cope with any distortion these processes bring to brand image
insights if it is known how these processes are ranked according to influence.
Potential Contributions
This study seeks to make the following substantive and methodological
contributions to the field of marketing:
1. Substantive: Provide empirically-based insights into how individuals reveal their
cognitive associations in groups, particularly blogs. Demonstrate the extent to
which social process theory explains the expression of thought in blog
communities, and thereby enhances our understanding of the usefulness (value)
of using blogs as a source of image insight.
2. Methodological: Demonstrate a novel integrative application of CRA with a type
of multidimensional scaling that uses a spring-based iterative algorithm, whereby
similar objects (in this case, bodies of text) are pulled together as dissimilar ones
are pushed apart until a stable equilibrium positioning is attained within a
multidimensional space. When these bodies of text have settled into stable
coordinates, thematic clusters are detected and measured to quantify diversity in
expressed thought.
12
CHAPTER II
OVERVIEW OF RELEVANT LITERATURE
A blog differs from a forum, newsgroup or general online community in that a
single author, or discrete set of authors, initiates the discussion. Most blogs permit
readers to add comments to posts and facilitate trackbacks, permanent linking to posts
by other blogs and websites. The collection of all blogs is often called the blogosphere.
Blogging pundits, such as Scoble and Israel (2006), hail the blog as an ideal forum
where organizations can converse with their market and constituents to gain insights into
how they are perceived. Since their widespread emergence after the Internet achieved
mainstream adoption, a number of marketing researchers have focused on virtual
communities in their work (e.g., Shoham 2004; Dholakia, Bagozzi and Pearo 2004);
however, there is a dearth of research studies that have used blogs as the research
context.
A discussion of the various kinds of image associations that consumers might
reveal in a blog follows in the next section.
Cognitive Associations
People tend to have very complex entwined cognitive associations with
organizations, brands and products. Govers and Schoormans (2005) make a distinction
between people’s perceptions of a brand and a product. Similarly, distinctions can be
made between an organization and a brand in its portfolio. Organization, product and
13
brand are levels in a hierarchy of consumer associations. While a consumer may have
different associations between these three levels, the nature of these associations is the
same in that they are all cognitive structures. While the literature reviewed in this
chapter tends to focus on one level of this hierarchy, the thoughts expressed generally
apply to all. Since this discussion is focused on cognitive structures in general, the words
organization, brand and product are used interchangeably with “brand” receiving the
most use.
Reputation
Brown et al. (2006) describe the perspectives from which an organization can be
viewed. Identity is the way an organization views itself. Intended image is the way an
organization would like to be perceived by outsiders. Construed image is the way an
organization thinks it is being perceived and reputation is the way it is actually
perceived. This taxonomy generally is inline with Gioia, Schultz and Corley (2000);
however, Gioia et al. made a distinction between image, the transient aspects of how an
organization is perceived, and reputation, the long-term perception. Brown et al. adopted
Brown and Dacin’s (1997) definition of reputation as organizational associations, a
“label for all the information about a company [organization in the broadest sense] that a
person holds” (p.104).
Brown et al. (2006) also made a distinction between reputation as an attribute of
an organization and reputation as the sum of all mental associations possessed by an
individual. This distinction is important as it explains why there is a need to distinguish
between the various organizational viewpoints: the way it is viewed by others is only
14
partially under an organization’s control. An organization is constantly communicating
through overt means such as advertising and through more subtle means such as its
products and its actions within the general community. Theorists have proposed various
models for how an organization’s message could be perceived differently than intended.
Shannon and Weaver (1949) introduced an encoding-decoding model where a message
is encoded by a sender, transmitted through a medium and decoded by a receiver into a
different message. They attribute a difference in comprehension to noise, random
interference, in the transmission medium. Later research, such as that reported by
Grayson (1998) attributes it to differences in the interpretive repertoire of the sender and
receiver. The sender encodes the message presuming that the receiver will not decode it
differently. Hall (1980) describes a circuit of communication whereby, through feedback
and clarification, a sender and receiver can converge on a shared meaning. The capacity
for reciprocal communication is one of the most attractive attributes of the blog.
Sabate and Puente (2003) concluded that a good corporate reputation results in
higher profits and higher profits result in a better reputation. Roberts and Dowling
(2002) decomposed reputation into financial (i.e., the company is good investment) and
general (the company is perceived as a good social actor) components. On this
foundation, Dowling (2006) concluded that companies can achieve superior financial
performance by investing in “being more profitable and in being perceived as good” (p.
135). Dowling observed that a good reputation aids corporate growth by making sales
easier and by making investors more willing to fund a company’s attempts at testing new
ventures. Obloj and Obloj (2006) noted that investments in increasing reputation are
15
done at the cost of investing in greater profitability. They concluded that superior profits
only accrue when there is a large difference between the reputations of competing firms.
Therefore, firms should strive only to be a close follower in reputation-building, not the
leader in their market.
Personality
In the brand identity prism proposed by Kapferer (1992), personality was
denoted as merely one facet of a body of interdependent associations that make up
image. Aaker (1997) took issue with this portrayal and proposed that it be regarded as a
separate construct: “the set of human characteristics associated to a brand.” Azoulay and
Kapferer (2003) argued that Aaker’s (1997) definition was too broad and not in keeping
with psychology’s definition of personality. They supported Kapferer’s (1992) model
that personality was merely an aspect of image and should only include those human
attributes that are appropriate to brands. As a result, they discarded cognition, skills and
abilities, gender, social class and ethnicity from the array of human characteristics that
can be associated with a brand.
Even though Azoulay and Kapferer (2003) advanced theoretical support for their
views, Freling and Forbes’ (2005) qualitative study of brand personality found that
consumers associate the full spectrum of human characteristics to brands. They ascribe
this to a natural human tendency to anthropomorphize objects with which they have a
relationship.
Belk (1988) noted that consumers make possessions a part of themselves and a
reflection of their identity. According to Belk, possessions serve a variety of self-
16
expressive purposes: they are a tangible link to one’s past as well as expressions of the
multiple levels of self such as family and groups one belongs to. It would not be
surprising to discover that the identity of a consumer may be so entwined with a
perceived brand image that they are unable to separate the two in their discourse.
Meaning
Brown, Kozinets and Sherry (2003) added the idea of brand meaning to the
pantheon of cognitive associations attached to brands. They described four classes of
meaning: allegory (a brand story), aura (core values), arcadia (idealized community)
and antinomy (paradox: old style, new technology).
Attitude
Petty, Unnava and Strathman (1991) noted that attitude is a relatively global and
enduring evaluation of a product. Ajzen (2001) differed by noting that attitude toward an
object may not be exclusively a global impression, but may be expressed on a series of
independent scales: bad-good, useful-useless, reliable-unreliable, etc. Attitude is relevant
to consumption because, as Ajzen (2001) notes “attitude predicts behavior,” people
purchase products they have a positive attitude toward. Howard and Sheth (1969)
position attitude as the mediator between choices and intention.
17
Relationship
Fournier (1998) distinguished a relationship from a one-time transaction and
mere habitual purchase from loyalty. Her definition of relationship also included the idea
that consumers and brands form partnerships of true interdependence where a brand is
typically attributed a human-like personality, much like the findings of Aaker (1997),
and the brand is perceived as an active partner in goal fulfillment. These goals can be
tangible/utilitarian and/or emotional as brands can facilitate the expression of self.
Fournier (1998) also points out that brand relationships exist in the context of other
relationships and may be entwined with them. She also stated that, much like human-to-
human relationships, brand relationships change over time. Arnould and Thompson
(2005) include the brand relationship in their taxonomy of consumer culture theory
because assigning human-like status to an intangible set of perceptions (i.e. a brand) is
an aspect of a constructed reality. John (1999) points out that children involve brands in
their socialization process and thereby create the foundations for brand relationships
early in life.
Knowledge and Expertise
Keller (2003) defined brand knowledge very broadly, including the intangible
aspects discussed here with the purely utilitarian attributes that compose most researcher
definitions. Mitchell and Dacin (1996) conceptualize knowledge as an awareness of
attributes and how they contribute to performance. Sujan (1985) includes the presence of
product categories in her definition, while Ratneshwar and Shocker (1991) unite
categorization with an appreciation of appropriate usage context. Expertise is
18
consistently presented as a high degree of knowledge combined with an ability to
properly apply it.
Community
Kleine, Kleine and Kernan (1993) observed that people use brands and products
in their identity-crafting processes. They noted that social connections are an important
part of that process because they allow a person to get feedback from knowledgeable
others about how well they are doing in their identity-crafting. As a result, the
knowledge of brand community is an important aspect of the mental associations that a
consumer has with brands. McAlexander, Schouten and Koening (2002) supplemented
that view by showing that association with a brand community increases a consumer’s
enjoyment of a product and provides a motivation for having a knowledge of available
communities.
Now that the primary image associations someone might reveal in a blog have
been described, in the next section the issue of corporate image management is
discussed.
Image Management
Figure 2.1 portrays a possible corporate image management process
incorporating blog insights. Companies have a self-image, the way they view
themselves, denoted their corporate identity. That identity is assumed to be the basis of
their intended image, the impression they would like to create in the minds of others.
Companies compare their intended image with a construed image, how they think others
view them. If the intended and construed images are too dissimilar, companies can try to
19
alter the way they are perceived by either tangible (i.e., through products and behavior)
or intangible (e.g., advertising and public relations) means. Consumers develop an
impression of a company from these image-building activities. That impression is
expressed in blog comments and subsequently read by corporate image monitoring. This
construed image is again compared with the image the company is trying to convey and
image building activities are once again adjusted to compensate for any mismatch. Blog
monitoring is thus an integral part of the image-building feedback loop.
FIGURE 2.1
Corporate Image Management with Blogs
20
The next section provides an overview of literature that supports examining blog
communications to detect consumer image associations. It then describes how these
associations can be detected and extracted.
Detecting and Extracting Cognitive Associations
As previously stated, this research project is concerned with investigating how
individual brand image cognitive associations are revealed in a group context. In order to
find patterns in how cognitive associations are revealed it is necessary to measure them
and the relationships between them.
Blogging, where an author posts an entry and the world is free to comment, is
similar to Olson and Muderrisoglu’s (1979) free elicitation procedure where
“respondents are free to say anything and everything that comes to mind when presented
with a stimulus probe cue.” They tested the stability of responses between two points in
time and found that 50-60% of the associations revealed in the first experiment also
appeared in the second. They also found that the order of mention of these associations
had a correspondence of about 0.35. This led them to conclude that free responses are a
reasonably consistent and reliable measure of thought structure.
Hutchinson (1983) demonstrated the use of a network-based model of cognitive
associations. He analyzed the transcripts of unconstrained free response from consumers
and product-category experts to construct a network of brand attributes connected by
directional ties (i.e. arrows) showing the order of recall (a temporal network). He built
his research on ideas introduced by a number of researchers who found that cognitive
associations are retrieved from memory in an order that reflects the way they are stored
21
(Bousfield and Cohen 1955; Bousfield, Sedgewick and Cohen 1954; Friendly 1979). At
the time when he proposed his network mapping of cognitive associations, two and three
dimensional visualizations based on data processed with multidimensional scaling
(Green, Wind and Claycamp 1975) was the state-of-the-art in modeling how “brand-
features” were represented in the mind of the consumer.
Ward and Reingen (1990) introduced another network-based model explaining
how a shared group cognitive structure emerges from individual structures and then, in a
kind of feedback mechanism, homogenized these individual cognitive structures. This
means that the knowledge or understanding of individuals becomes shared group
knowledge as individuals express their thoughts. Hearing the thoughts of others cause
individuals to modify their personal knowledge, making it more like the knowledge
shared by the whole group. They used this model to explain the process of how a group
arrives at a consensus-based consumption decision. Their work was heavily based on
theories of group polarization, such as those advanced by Burnstein and Sentis (1981).
Groups tend to be attracted to the “initially favored alternative”, that is, the first
alternative that is well explained and supported by logical arguments. Chandrashekaran,
Walker, Ward and Reingen (1996) expanded this model by demonstrating how decisions
could be predicted and not merely explained.
It should be noted that all this prior research confined cognitive association maps
to those involving the purely utilitarian attributes of products, believed at the time to be
most relevant to the purchase decision. As described, consumers associate a much richer
set of intangible and non-fungible attributes to organizations, brands and products. As
22
Keller (2003) points out, the mapping of all these cognitive associations has not been
explored:
An important future research challenge … is to develop holistic
perspectives toward brand knowledge that would encompass the full
range of all the different kinds of information involved, for example,
approaches to create and apply detailed mental maps for brands. An ideal
mental map representation would be a blueprint of brand knowledge, as
comprehensive while also as parsimonious as possible, that would
provide the necessary breath and depth of understanding of consumer
behavior and marketing activity. (p. 596)
Even though early research embarked on a very promising trajectory it seems to
have stalled, possibly waiting for more advanced methodologies to be developed. The
paragraphs to follow set the stage for introducing one such new methodology.
Content Analysis
Among the many definitions of content analysis summarized in Kassarjian
(1977), the one provided by Kerlinger (1964) most closely fits the objectives of this
study:
Content analysis, while certainly a method of analysis, is more than that.
It is … a method of observation. Instead of observing people’s behavior
directly, or asking them to respond to scales, or interviewing them, the
investigator takes the communications that people have produced and
asks questions of the communications. (p. 544)
Most of the current methods of content analysis are unchanged from the ones
described by Kassarjian (1977). Methodologies tend to be of two types: manifest and
latent. Manifest methods tend to be word use frequency counting systems that cannot
capture the way words are used in context, only that some words are used more often
than others. Latent methods generally employ a group of human interpreters to overcome
23
the deficiencies of manifest methods. However, latent methods are often characterized
by inter-rater reliability problems.
Thompson (1997) describes a “hermeneutical framework for deriving marketing
insights” from consumer-generated text that is particularly germane to this study. He
uses the word hermeneutic to denote a situation where the consumer reflects on the role
products play in their present and past circumstances in the context of the broader
culture in which they live. The nature of a blog encourages commenting consumers to
write at length about an issue that interests them. This permits them sufficient scope for
reflection and elaboration. The ongoing communal nature of the blog encourages repeat
contribution, allowing them to further elaborate on issues they want to discuss as time
and reflection allows them to refine their thinking.
Centering Resonance Analysis
Centering Resonance Analysis (CRA), a method of content analysis proposed by
Corman et al. (2002), turns a body of text into a network of nouns and adjectives. The
words that have the most effect in giving the text coherence are then found by finding
the words with the highest betweenness centrality, or simply betweenness. These are
words that are important because they connect other words together. Figure 2.2 shows
part of the word network composed from a comment posted to GM’s Fastlane blog. It is
apparent that the word “oil” has high betweenness because it connects many other words
together. Corman et al. (2002) postulate that these word networks provide insights into
the way knowledge is structured in the mind of the writer. The words with high
betweenness represent the most salient cognitive associations. A quantitative measure of
24
a word’s betweenness is called its influence. Two separate bodies of text can be
compared on the basis of commonalities in the influence of the same words used. A
quantitative measure of this similarity is called resonance.
FIGURE 2.2
Betweenness Centrality in Text
CRA operates in a manner consistent with Thompson’s (1997), framework
because it models a whole body of text as one coherent network and seeks to detect the
role specific words play in that network of meaning. The ability to measure the
resonance between separate texts allows the researcher to construct an integrative
interpretation. This study proceeds on the premise that the most influential words in the
25
comments are also the most salient organizational, brand and product associations
elicited by the author’s stimulus posting.
Having described the means of extracting cognitive associations from a body of
text, the next section discusses how the many cognitive associations that consumers
reveal can be organized to allow the detection of patterns among them.
Mapping Cognitive Associations
Multidimensional Scaling
Multidimensional scaling (MDS) (Green, Wind and Claycamp 1975) has become
an important tool of marketing research as it is typically used to create two or three
dimensional visualizations of data that leverage a natural human facility for identifying
patterns. Even though many datasets reflect more latent dimensions than can be reduced
effectively to two or three dimensional visual models without too great a loss of
information, there is no need to abandon MDS in such situations. Strictly speaking,
MDS is not a technique for data visualization; it is a means of converting pair-wise
similarity measures expressed as distances (like one minus the resonance between two
bodies of text) into coordinates in a space of any dimensionality. Matrix manipulation
and spring models (described below) are the two most common ways of performing
MDS. Spring models are usually favored when a large number of distances are used
because the manipulation of large matrices is time consuming and resource intensive
when performed on personal computers. Since the use of spring model MDS is unusual
in marketing, the underlying principles are described here. However, first an even more
26
unusual means of performing MDS using neural networks is described because the
founders of that methodology have used it in situations similar to that of this study.
Woelfel and Fink (1980) introduced Galileo, a computer program that extracts
“objects of cognition” from a body of text and plots them in a multidimensional space
“whose properties are determined by the patterns of interrelations among the objects” (p.
8). Galileo embodies the research findings of Barnett and Woelfel (1979) who
documented critical problems with reducing psychological measurements to the minimal
set of eigenvalue-based dimensions commonly used to define a multidimensional scaling
space. Woelfel and Fink (1980) still claim (Woelfel 2007) to be the only researchers who
have successfully mapped the cognitive content described in a body of text to a multi-
dimensional space. In order to do so, they employed a neural network to translate such
textual content to coordinates. Typically, users of two and three dimensional scaling
have endeavored to assign labels to the axes of the space that denote attributes on which
objects (whether products, customers or cognitive objects) vary. Barnett and Woelfel
(1979) point out that in high dimensional spaces, the axes cannot be labeled and
psychological constructs are points within the space.
27
Spring Models in Multidimensional Scaling
Kamada and Kawai (1989) devised the algorithm used to perform MDS with
spring models. Their algorithm models each data point as connected to every other data
point by a spring tensioned according to some measure of similarity (such as CRA
resonance). Through an iterative mechanism, similar data points are pulled together as
dissimilar ones are pushed apart until an equilibrium configuration is attained. If that
methodology were used in this study, representing blog entries and comments as spring-
connected objects, once the objects self-organized into stable positions within multi-
dimensional space, thematic clusters could be detected using methods of inherent
classification such as those discussed below. Thematic clusters can then be measured in
the following ways to indicate diversity in expressed thought:
1. The mean number of clusters across blog entries can be a blog-wide measure of
cognitive diversity, while the number of clusters within a blog entry can indicate
the degree of cognitive diversity within a single blog entry conversation.
2. Radius and density: Each cluster has a center of mass (also called a centroid or
barycenter). The mean and standard deviation of the distance between every
cognitive object and the centroid are measures of a cluster’s radius and density,
respectively.
3. Proportion of outliers: The mean and standard deviation of the distance between
every cognitive object and its cluster centroid can also be used to identify
outliers, cognitive objects located far from a cluster’s center.
28
It should be apparent from the above description that the CRA and spring model
multidimensional scaling algorithms are particularly appropriate for automation as they
make no qualitative determinations that imitate human judgment. There is, therefore, no
need to employ multiple human referees to assess the reliability of CRA or spring model
multidimensional scaling as the mapping of the same body of text to multidimensional
space is certain to have identical results regardless of the number of trials.
Inherent Classification
Most clustering algorithms, such as K-Means and Expectation Maximization
(EM), require that the researcher specify the number of clusters the algorithm is to find.
However, in this research study, it is necessary to know what clusters are naturally
present. One popular clustering algorithm that does not require the number of clusters to
be specified beforehand was developed by Chiu et al. (2001). Chiu et al. described an
entropy-based hierarchical clustering algorithm that implements agglomerative
hierarchical clustering in two stages:
1. Pre-cluster the data points into a large number of sub-clusters (Chui et al. calls
these “dense regions”).
2. Cluster the resulting sub-clusters into a smaller “optimal” number of clusters.
In the first phase, Chiu et al. (2001) use Zhang et al’s (1996) BIRCH method of
constructing a cluster feature (CF) tree. Each CF consists of a finite number of data
points that are within a certain radius of each other. The maximum number of data points
per CF and the radius must be selected prior to building the tree. As the dataset is
29
traversed, data points are added to CFs, new CFs are created and existing CFs are split
until every data point in the dataset is accommodated.
Prior to the second phase, the “optimal” number of clusters is determined in its
own two-step procedure:
1. Calculate a coarse estimate of the optimal number of clusters using Bayesian
Information Criterion (BIC). BIC is a likelihood-based metric that decreases as
the number of clusters increases to a certain point and then increases beyond that
point. The aim of this first step is to find the number of clusters at the BIC
minimum. Chiu et al. (2001) note that since the BIC estimate almost always
overshoots the true optimum number of clusters, another step is needed to refine
the estimated optimum.
2. Simulate the merging of all the sub-clusters and calculate the ratio of inter-cluster
distance and the number of clusters after each merge. A big jump in the change
in this ratio from one merge to another usually indicates that merge should not
have occurred. This process can also be done by plotting the number of clusters
versus classification error and then locating the “knee” of this curve. This method
is sometimes criticized for finding the knee locally by looking at pairs of adjacent
points rather than at the whole curve. It is however, a simple method to use.
After the optimal number of clusters has been determined, the CF tree is
collapsed into that number of clusters based on the combining of sub-clusters in close
proximity to each other. Proximity is measured as a log-likelihood distance related to the
30
decrease in log-likelihood as the two sub-clusters are combined into one cluster. The
distance between sub-clusters w and z is calculated as defined in equation 2.1.
(2.1) ><−+= zwzwzwd ,),( ξξξ
In equation 2.1, <w, z> denotes the index of the combined cluster and ξ is the entropy, or
information homogeneity, of a cluster. The more homogeneous a cluster with respect to
the distribution of values within it, the lower its entropy. This clustering method
therefore tries to join sub-clusters where the difference between their separate entropies
and their joined entropy is small.
Now that a means of organizing and analyzing the relationships between
cognitive associations has been described, the literature that offers explanations for these
relationships and any patterns detected is reviewed next.
Social Process Theory
Simmel introduced the idea that all social phenomena are the collective result of
transactions between individuals (Simmel and Levine 1972). He observed that there is
no universal law or force compelling any social act from the top down, it is solely
created from the bottom up by individuals acting under their own volition. Cederman
(2005) sees Simmel as the progenitor of the science of complexity, the search for simple
rules (also called generative rules) that underlie complex phenomena. Phelan (2001)
differentiates between traditional and complexity science:
Traditional science seeks direct causal relations between elements in the
universe whereas complexity theory drops down a level to explain the
31
rules that govern the interactions between lower-order elements that in
aggregate create emergent properties in higher-level systems. (p. 128)
Cederman (2005) builds on this concept in defining social process theory:
The sociological process approach starts with an observed social
phenomenon, whether unique or ubiquitous, and then postulates a process
constituted by the operation of mechanisms that together generate the
phenomenon in question. (p. 867)
The processes that operate together to generate a phenomenon are often called
social forms.
The blog is a prime example of an arena dominated by the dynamic of
complexity: the blog author can post content but cannot compel response. Readers are
moved to respond in a manner reflecting the diversity of ways they interpret the blog
entry, possibly inspiring a cascade of responses from more readers in a cycle that repeats
until readers no longer have anything to write. In the remainder of the chapter, six social
forms that previous research has found to govern collective action are discussed. These
six complex processes do not constitute an exhaustive list, but are merely the processes
that previous research suggests might be most operant and important to communities in
the blogosphere.
The Echo Chamber
Scoble and Israel’s (2006) “echo chamber” effect refers to the illusion of a
vibrant community that frequent communication between a few parties can create:
Blogging can fool you. You may think you are conversing with the world,
when it’s just a few people talking frequently, back and forth to each
other, creating the illusion of amplification. The echo chamber can
deceive a business into thinking it is either more widely successful or
further off the mark than it is in reality, because a few people are making
a lot of noise. (p. 134)
32
Three similar phenomena have been addressed in academic research under the terms:
cultural tribalism, groupthink and group polarization. Kitchin (1998) described cultural
tribalism in this way:
… communities based upon interests and not localities might well reduce
diversity and narrow spheres of influence, as like will only be
communicating with like. As such, rather than providing a better
alternative to real-world communities cyberspace leads to dysfunctional
on-line communities … (p. 90)
Cultural tribalism is thus portrayed as the ultimate equilibrium condition of all online
communities. Since the cost of trial and switching are low, people will sample a large
number of communities and migrate to the ones wherein they feel most at home, those
where they hear enough of what they want to hear to feel cognitively at ease. Such
groups are likely to be small, due to an intolerance of dissent, albeit close-knit.
Conversation may be lively, but always among the same people, sharing a limited range
of thoughts. Therefore, this migration to comfortable cognitive spaces causes cognitive
diversity to be sacrificed. If cultural tribalism is as strong an influence as Kitchin (1998)
opines, then communities characterized by diverse thought are likely to diminish over
time with high cognitive diversity being only temporary.
33
FIGURE 2.3
The Power-Law Distribution of Attention in the Blogosphere
This argument is supported by Shirky (2003), whose meta-analysis cites the
findings of several researchers who observed from different perspectives that attention
metrics (e.g., numbers of readers, comments and trackbacks) in the blogosphere tend to
follow a power-law distribution similar to that depicted in Figure 2.3. Only a few very
popular blogs attract the attention of a large number of people while most blogs,
occupying the long tail of the distribution, attract a niche of people whose unique
preferences closely match the blog’s content.
34
The discussion thus far might well lead one to wonder whether the
mechanism that underlies cultural tribalism is solely one of random drift to
centers of thought sameness. Wallace (1999) offers another perspective as he
describes a different situation that will lead to low diversity in ideas expressed:
Each person can share what he or she knows with the others, making the
whole at least equal to the sum of the parts. Unfortunately, this is often
not what happens…. As polarization gets underway, the group members
become more reluctant to bring up items of information they have about
the subject that might contradict the emerging group consensus. The
result is a biased discussion in which the group has no opportunity to
consider all the facts, because the members are not bringing them up…
Each item they contribute would thus reinforce the march toward group
consensus rather than add complications and fuel debate. (pp. 81-82)
Wallace (1999) uses the term “polarization,” but is really describing a similar
phenomenon called groupthink. The mechanisms of the two phenomena are similar but
the outcomes differ. In groupthink, a group’s desire for unanimity causes people to
withhold information that might contradict what they perceive to be a growing group
consensus (Janus 1972). In group polarization, people gradually come to advocate
extreme versions of their personal beliefs causing the group to divide into more
homogeneous sub-groups along well-demarcated ideological lines (Moscovici and
Zavalloni 1969). The two processes differ in their details, but both involve individuals
compromising their personal knowledge or beliefs to achieve a form of collective unity.
It is possible therefore that cultural tribalism is a superficial condition that does not truly
indicate the diversity of thought among its members, it may be the result of a deliberate
seeking of collective unity among otherwise diverse thinkers.
35
However, Matz and Wood (2005) found that heterogeneous attitudes create
dissonance or tension and discomfort between members of a group. They found that the
level of such discomfort was proportional to the numerical minority status of those with
atypical views. The discomfort is partly relieved if the minority view-holders felt free to
affirm their attitudes without pressure to conform. But, true relief of discomfort was only
achieved if the minority could persuade the majority of their error, the majority could
present more convincing support for their position, or if the minority could leave and
join a more compatible group. It should be concluded then that compromising personal
values to achieve collective unity is not a sustainable situation in the blog context as
unity is not a forced necessity. The migratory aspect of cultural tribalism, given its ease
among Internet blogs, seems to be the result of a natural, unavoidable coping
mechanism.
Flocking Theory
Reynolds (1987) proposed flocking theory as a computational model that
explains how the coordinated movement of a group can emerge from individuals making
decisions based on personal information. He discovered that by using three simple
“steering behaviors,” coordination would emerge without any explicit management
activity: separation (“steer to avoid crowding local flockmates”), alignment (“steer
toward the average heading of local flockmates”), and cohesion (“steer toward the
average position of local flockmates”). These steering behaviors are really just heuristic
components of the more complicated calculation of which direction an individual should
move to be in its desired location, relative to the group, one unit of time in the future.
36
Although flocking theory was developed as a solution to modeling the behavior
of flocking birds and animals in computer graphics, it has been extensively investigated
in a variety of academic disciplines and mathematically modeled by physicists, Toner
and Tu (1998). Rosen (2002) proposed that flocking theory was a good explanation for
self-organization in human social systems. He proposed that communication was the
mechanism of cohesion in human society where a social network of individuals shares
access to a collective body of knowledge that acts as a “roadmap” for coordinated action
with little centralized control.
Rosen (2002) based his model on multiple literature streams. Simmel and Levine
(1972) state that for social relationships to occur “the personalities must not emphasize
themselves too individually … with too much abandon and aggressiveness.” Eisenberg
and Phillips (1990) proposed that community cohesion is always a balance between
autonomy and interdependence, corresponding to Reynold’s (1987) separation and
cohesion steering behaviors. Rosen (2002) concludes that to some extent uniformity and
common interest is essential to flock maintenance and that individuals must sacrifice
some autonomy to keep group acceptance.
From the discussion thus far, flocking and the unity-seeking compromises of
groupthink and group polarization may seem not to be distinct concepts. The important
difference is that people who engage in groupthink and group polarization are motivated
by a perceived need for collective unity, while people who engage in flocking
(“flockers”) are motivated by a personal need for social belonging. Additionally, in the
blogosphere, nothing constrains a flocker from engaging in a cultural tribalism-style
37
migration to another blog if the discomfort felt from compromise exceeds the value
gained from social acceptance. So the flocker’s compromise probably causes less
dissonance than that felt by the migrating tribalist.
Threshold Models of Collective Behavior
Granovetter (1978) criticized previous research on crowd behavior for being too
accepting of a universal model of individual compliance with social norms. He used as
an example the situation of a riot where individuals who would never start a riot would
participate nonetheless if sufficient other individuals joined in. He proposed that
everyone had a threshold for such behavior and that if the diversity of individual
thresholds were uniformly present in a crowd, a riot, once started, would propagate
throughout the group getting everyone to participate. This model can be readily applied
to a blog, where a few readers might have something they would like to say but are too
shy to be the first to comment. Once some people who are less shy express their
thoughts, the shyer will feel comfortable enough to post their comments.
Memetics
Marsden (1998) called memetics the study of mind viruses, ideas that are highly
contagious and quickly spread through a population. The term “memetics” is derived
from meme, a word introduced by Dawkins (1976) in his classic book The Selfish Gene.
Dawkins (1976) was primarily concerned with arguing that most human behavior is
directed toward genetic propagation. However, as Blackmore (2001) observed, he
wanted to show that the core processes of evolution (mutation, propagation and natural
selection) are not confined to biological genetics but also to other replicators such as
38
ideas. Through memes or systems of related memes (what Blackmore called memesets),
humanity passes certain cultural practices from generation to generation because these
practices increase survivability. Marsden (1998) described a three stage process where
acquired ideas (“you get an idea”) become ideology (“the idea gets you”) resulting in
specific behavior. It is readily apparent that this mechanism is superior to forcing people
to behave in a certain way because when the force stops the behavior will probably stop.
However, if a person can be persuaded as to the value of a certain behavior, then that
behavior is much more likely to be continued indefinitely. Just as not everyone is equally
susceptible to catching a biological virus, there are various degrees of resistance and
immunity to memes. Dawkins (2006) described two strategies that have evolved to
maximize the success of meme propagation: intense indoctrination and mass
dissemination. During intense indoctrination, the target person or population is subjected
to the high repetition of a meme, or memeset, in an attempt to overwhelm thresholds of
resistance. Indoctrination, in its attempt to suppress the opposing influence of critical
thinking, often involves the exclusion or ridicule of competing memes. In mass
dissemination, the targets are the people with the lowest barriers to adoption, so the idea
is spread as widely as possible to increase the probability of reaching people with a low
threshold of resistance.
It may seem from the discussion thus far that there is little difference between
contexts influenced by intense indoctrination and those influenced by cultural tribalism
and flocking. Cultural tribalism and flocking are purely grassroots phenomena, while
indoctrination is driven by top-down intention to persuade others. Ultimately, to be
39
successful, efforts at indoctrination must inspire grassroots support. People most
susceptible to, and accepting of, the indoctrinated meme will cluster (cultural tribalism)
while others, motivated to be accepted by the core, flock around the periphery.
Need-for-Cognition
Thus far in the discussion, two innate motivations have been discussed along
with their resultant behaviors: the desire for uncompromised self-expression, revealed in
cultural tribalism, and the desire for acceptance, revealed in flocking. It is apparent that
neither of these two motivations contributes to the creation of diverse thought within a
blog as both are associated with situations where diversity of thought will be either
reduced or constrained. For blogs to be centers of diverse thought, influences must be
present that tend to increase the expression of diverse perspectives. As already stated, the
focus of this study is detecting diversity in attitudes toward organizations, brands and
products. It is proposed that the best place to see what attitudes are present is one where
attitudes are formed and modified. The blog context is likely to be such a place as blog
authors, even without exercising such extreme methods as indoctrination, are likely to
use persuasive language intended to affect attitudes. Petty and Cacioppo (1986)
proposed one of the most accepted models of attitude change: the Elaboration
Likelihood Model (ELM). The concept of an elaboration continuum, where thought
among targets of persuasion varies from low to high, is central to the ELM. Part of what
determines the level of thought exercised by targets of persuasion is their innate need-
for-cognition, an enjoyment of cognitively demanding tasks. A community of blog
commenters that reflects the full diversity of the market should exhibit the full spectrum
40
of the elaboration continuum in their commenting activity. Commenters at higher need-
for-cognition levels should be a grassroots influence that increases the level of thought
diversity expressed in a blog.
Thus far in this discussion, need-for-cognition is portrayed as an internal
motivation that inspires thinking. Neither the psychological nor the marketing literature
connects need-for-cognition with social behavior. However, past research has described
the phenomenon of sensemaking as an internal activity that inspires collaborative social
activity. In the next section, sensemaking is described and need-for-cognition is
suggested to be its underlying motivation.
Need-for-Cognition and Sensemaking
Sensemaking, as the word implies, is an act of making sense of circumstances. It
has already been noted that communicated messages are often perceived differently than
intended by a sender and differently again among receivers because of a noisy
transmission medium or because of differences in the interpretive repertoire of senders
and receivers. Thus, it should be no wonder that blog entries and comments might be a
source of reader confusion. The conceptualization of sensemaking in the blogosphere
used in this dissertation is based on a framework developed by Weick, Sutcliffe and
Obstfeld (2005) to explain how sensemaking occurs, and is a formative process, in
organizations. The association of organizing with blogging is particularly apt because
both are contexts where individuals react to their understanding of current conditions
and thereby shape the evolution of a community. Weick et al. describe sensemaking as
an ongoing individual process of retrospection where a person asks “what is going on
41
here?” and then uses the explanation they consider most plausible as a basis for action. It
seems reasonable to propose that the depth of such questioning and subsequent inquiry
should be related to an individual’s need-for-cognition. The action resulting from an
individual’s sensemaking changes the context, inspiring others to enter the sensemaking
process to update their understanding of what’s going on before they perform their own
actions. Hence, individuals enter and leave the sensemaking process in an ongoing cycle
driven by the actions of community members. Weick et al. see the sensemaking process
as a response to surprising or unexpected actions; however this study sees the element of
surprise as only affecting the degree of sensemaking an individual undergoes in response
to change. Since there is always a mismatch between how an act is intended and
perceived, all action contains an element of the unexpected.
Weick et al. see sensemaking as starting with a noticing of something “at
variance with the normal” (also described as “disruptive ambiguity”) that is then
investigated and “labeled” so as to become part of the common currency for
communicational exchange. The investigation begins with a personal retrospective
search of past experience in the person who first notices the novel event. The result of
that personal search is the initial labeling. Weick et al. see sensemaking as always a
social process where the initial labeling facilitates discussion with others where the
sensemaking continues and propagates, possibly resulting in a change of the labeling:
“thinking is acted out conversationally.” Behind the sensemaking is always a
consideration of response: “what should I do?” Action and talk thus become a cycle.
Even though every element of conversation in a blog is preserved, not every thought
42
expressed becomes the impetus for new thought expression. The more unexpected the
thought, the more it is likely to become part of a chain of collective learning.
In the marketing literature, Rosa et al. (1999) show how product markets are
knowledge structures that are socially constructed through a collaborative sensemaking
process among producers and consumers. In this process, consumers and producers tell
each other stories in published media: the producers convey what the products can do;
the consumers reveal what they need and their experiences with products. Through the
process, consumers and producers converge on a common understanding of what a
product category can and should do. The blog should be an ideal platform for the
sensemaking process because it is publicly visible and designed to be an interactive
conversational tool.
Cognitive Diversity
In the preceding sections on cultural tribalism, flocking, memetics and need-for-
cognition, their relevance to cognitive diversity (the diversity of thought expressed in a
blog) has been the focus of discussion. For this reason, the focus of this study is on
cognitive diversity as a construct. This is not the first study to use cognitive diversity as
a focal construct. Recent works by Page (2007) and Surowiecki (2004) extol the virtues
of cognitive diversity in groups of all kinds: diverse thinking leads to better collaborative
outcomes. Page (2007) recognizes a dual nature to diversity in cognition: diversity in
preferences and diversity in thinking styles, or as he calls them, cognitive toolboxes.
Page also notes there is some degree of interdependence between preferences and
43
toolboxes since people develop the thinking skills needed to satisfy their preferences and
then change their preferences based on their new ways of thinking (pp. 285-296).
Reciprocity and Game Theory
A blog is a public good, depending on the participation of a community for
value, but not diminished in value by the number of readers. In a blog, where the author
posts (a costly effort) in anticipation of response (a payoff), there is always uncertainty
about the amount of response the author will receive from any post in return for his cost
of posting. This is an example of a trust-reciprocity scenario, shown by Pereira, Silva
and Andrade e Silva (2006) to be robust in certain situations:
… reciprocity is a pattern of behavior that may be considered ‘robust’
because it emerges in specific environments where the marginal costs of
such behavior are reasonable, but not prohibitive. (p. 420)
Since the unique feature of a blog is that a single author controls the discussion,
each blog entry is like a sequential game where the author moves first. As already stated,
this first move is in expectation that the value received from all the resultant community
contributions will exceed the cost of posting. If Pereira et al’s conclusion is applied to
blogs, then it can be supposed that if the value of the author’s post exceeds the cost of
reciprocal response, community members will reply under the pressure of a sense of
obligation to reciprocate equal to the difference between value received from the post,
and the effort expended to contribute a comment. The author’s original post plus the
cascade of member contributions will continue to raise the value received by the
community, increasing the level of obligation felt by those that have not contributed,
inspiring more contribution as thresholds of tolerance for withheld reciprocation (i.e.
44
cost of reciprocation) are exceeded. Some community members may have very high
levels of tolerance or costs of reciprocation and be free riders in any specific game.
While their receipt of benefits does not diminish the value of existing content to the rest
of the community, it does cause the maximum value of content to peak at less than what
could result from full participation. While the author’s post might be seen as an act of
altruism, it is rational to assume the author will eventually stop trying to initiate a
response if repeated attempts generated less return than the cost of posting. This model’s
“thresholds of tolerance” is similar to the type of model proposed by Granovetter (1978).
Summary
This chapter provides a review of the principal streams of literature relevant to
this research study. Brand image is encoded in the minds of consumers in a variety of
ways. Stimuli, similar to that provided by a blog author’s entry, evoke free-form
responses from consumers that reveal their brand associations. These responses can be
mapped to a multidimensional space where similarity between cognitive objects can be
assessed by looking at their spatial proximity. Various social processes have been
described that purport to explain patterns in the spatial proximity of cognitive objects,
including social process theory, which ascribes these patterns to the interaction of many
social processes. This review suggests that social process theory offers a credible
explanation for the way people contribute to blogs and the diversity of thought they
express. The next chapter builds on this conclusion and proposes a conceptual model and
hypotheses to test it.
45
CHAPTER III
CONCEPTUAL MODEL AND HYPOTHESES
As previously stated, this research study investigates the extent to which postings
and comments in blogs reflect a company’s image in the overall market. This research
study seeks to lay a foundation for answering the above by testing for the effect of six
social processes (or forms), identified in prior research and detailed in Chapter II, that
are known to have an effect on the diversity of thought expressed in a group forum.
First, this chapter presents a discussion on this study’s dependent variable, a
latent variable construct representing cognitive diversity, that is, diversity of thought
expression. Next, six latent independent variable constructs are described. Each
represents one of the social processes described in Chapter II. Two conceptual models,
hypotheses and conceptual support for the hypotheses are presented.
Cognitive Diversity
Once multidimensional scaling has located blog entries and their associated
comments into stable positions within cognitive space, thematic clusters can be detected.
In Chapter II, four cognitive diversity metrics were introduced from measures of
thematic clusters: the number of clusters associated with a blog entry, their mean radius
and density, and the percentage of outliers. As depicted in Figure 3.1, these metrics can
be used as indicator variables for the diversity of thought (i.e., cognitive diversity) latent
variable. The influences on diversity of thought are all formative effects and are defined
as independent variables in the sections to follow. As stated in Chapter I, blog authors
46
write posts or entries that prompt commenters to write replies. Since it is the relative
locations of comments to each other and their associated blog entry that are the basis for
this study’s measures of cognitive diversity and the constructs hypothesized to affect it,
the blog entry is the unit of analysis, unless otherwise noted.
FIGURE 3.1
Cognitive Diversity Latent Variable
47
Cultural Tribalism, Need-for-Cognition and Cognitive Diversity
In Chapter II, cultural tribalism was described as a process whereby like-minded
people are brought together by a desire to feel validated by the exercise of
uncompromised self-expression. Since this is generally only possible among people who
agree, it results in the creation of contexts were members only hear their own limited
ideas and perspectives echoed by each other. Need-for-cognition was described as a
competing motivational influence where people who like intense thinking are attracted
into the blog conversation to engage in high intensity cognitive tasks (like sensemaking)
and thereby expand the diversity of thought expressed. As explained more fully next,
this study conceives of these two competing motivations as occupying opposite
quadrants of a conceptual space defined by variations in individual thought diversity
(i.e., the variety across an individual’s past and current idea expressions) and collective
thought diversity (i.e., current variety across different individuals) as shown in Figure
3.2.
48
FIGURE 3.2
Cultural Tribalism and Need-for-Cognition as Opposing Motivations
In Chapter I, it was mentioned that blog monitoring can be part of a feedback
loop, allowing companies to monitor and adjust their self presentation. As such, blog
monitoring facilitates organizational learning. March (1991) argues that cognitive
diversity is the key to such knowledge development because it results in new alternatives
being applied to problem solving rather than a constant recycling of old solutions. March
opines that cognitive diversity is achieved by the constant influx of new voices; these
new people will not have developed enough attachments with or within the group to be
overly concerned with achieving cohesion or forming alliances. Since cultural tribalism
49
causes people to form long term associations with communities wherein they feel most
comfortable, it is reflected in the mean longevity, or mean time between the first
comment and the current comment, of commenters to the current blog entry (“Mean
Commenter Longevity”) and the number of commenters who commented on the last
blog entry (“% Repeat Contributors (T-1)”). The latter indicator captures a recency
effect; that is, it is assumed that if a person commented on the last blog entry, then they
are more likely to comment on the current blog entry and thus lower the diversity in
thought expression for the whole community. This assumption is based on Holme’s
(2003) finding that ties to virtual communities decay over time with decreasing contact.
A recent comment is evidence of current membership and therefore a strong predictor of
further comments in the immediate future.
However, long term associations are not sufficient to distinguish cultural
tribalism from the other influences that affect cognitive diversity. Commenters that
receive value from satisfying their need-for-cognition can also be expected to be regular
commenters. As described in Chapter II, cultural tribalism is predicted to cause closer-
knit relationships between people who express similar views. As a result, Figure 3.3 also
depicts cultural tribalism as reflected by a negative relationship to: (1) the mean entry-to-
comment and comment-to-comment cognitive distance (denoted “Mean Collective
Thought Separation” in Figure 3.3), (2) the standard deviation of entry-to-comment and
comment-to-comment cognitive distance across blog entries (denoted “Stdev Collective
Thought Separation”), (3) the mean cognitive distance between an individual’s
comments across blog entries (denoted “Mean Individual Thought Separation”), and (4)
50
the standard deviation of the pair wise cognitive distances between an individual’s
comments across blog entries (denoted “Stdev Individual Thought Separation”). Need-
for-Cognition is depicted in Figure 3.3 as reflected by positive relationships to the same
indicators. Thus,
H1: Cultural tribalism is negatively related to cognitive diversity.
H2: Need-for-Cognition is positively related to cognitive diversity.
Hypotheses 1 and 2 are reflectively modeled in Figure 3.3. In Figure 3.3, the unit
of analysis is the blog entry.
Blog entry comments primarily influenced by cultural tribalism can be
distinguished from those primarily influenced by need-for-cognition by a discriminant
model (equation 3.1) that differentiates on the basis of diversity of thought expressed
using similar metrics to those depicted in Figure 3.3.
(3.1) 0 1 2 3 4j C C I IZ d dβ β β σ β β σ= + + + +
51
FIGURE 3.3
Cultural Tribalism / Need-for-Cognition Reflective Sub-Models
In equation 3.1, the unit of analysis is the blog entry comment, Zj is the
discriminant Z score of the discriminant function j, Cd is the mean entry-to-comment
and comment-to-comment cognitive distance across comments to the current blog entry
(a limited form of the metric denoted “Mean Collective Thought Separation” in Figure
3.3), σC is the standard deviation of entry-to-comment and comment-to-comment
cognitive distance across comments to the current blog entry (a limited form of the
metric denoted “Stdev Collective Thought Separation” in Figure 3.3), Id is the mean
comment-to-comment cognitive distance for the current commenter across blog entries
52
(a limited form of the metric denoted “Mean Individual Thought Separation” in Figure
3.3) and σI is the standard deviation of comment-to-comment cognitive distance for the
current commenter across blog entries (a limited form of the metric denoted “Stdev
Individual Thought Separation” in Figure 3.3).
The parameters of equation 3.1 can be used to distinguish between contexts
primarily motivated by cultural tribalism and those motivated by need-for-cognition. To
conceptually distinguish between cultural tribalism and need-for-cognition, the
parameters of equation 3.1 can be divided into two groups to create sub-factors that
represent individual thought diversity (mean and standard deviation in individual thought
separation) and collective thought diversity (mean and standard deviation in collective
thought separation). These sub-factors are used as axes of the perceptual map shown in
Figure 3.2. If blog entry comments are plotted on this perceptual map, their type should
be readily distinguishable by their presence in one of the marked quadrants. Since the
unit of analysis of equation 3.1 is the comment, comments must be added to Figure 3.2
one-by-one in the order they were created. Individual and collective thought diversity
coordinates would be calculated at each step using the comment being added, and its
thematic relationship to the comments and blog entry already present.
Threshold Models of Collective Behavior and Cognitive Diversity
In Chapter II, Granovetter’s (1978) model of a generative process was described
as a process that builds on itself. Granovetter used his model to elaborate an example
scenario that assumes everyone has a threshold for violent behavior. If the full variety of
individual thresholds were uniformly distributed within a crowd, a riot, once started,
53
would propagate throughout the group getting everyone to participate. The idea behind
this process is used in this study to propose a mechanism for the emergence of divergent
themes within the comments to a blog entry. In this situation, the existence of a class of
commenters, denoted idea seeders is predicted. Idea seeders are the riot-starters of a
blog. They are among the most uninhibited, and therefore the most prolific, commenters.
They are willing to express ideas on the most diverse array of themes. Just as riot
participation spreads from the originators to the rest of a crowd in a cascading fashion,
idea seeders are consistent starters of new thematic clusters; their willingness to express
themselves emboldens others with similar ideas to write comments, cascading through
the readership until all like thoughts are expressed. As the number of commenters
increase, the probability of having an idea seeder among them likewise increases. Thus,
H3: Idea seeding is positively related to cognitive diversity.
Flocking Theory and Cognitive Diversity
In Chapter II it was noted that Rosen’s (2002) proposal that flocking theory, a
model explaining how the coordinated movement of a group can emerge from
individuals making decisions based on personal information, was a good explanation for
self-organization in human social systems. Rosen concluded that, to some extent,
uniformity and common interest is essential to flock maintenance and that individuals
must sacrifice some autonomy to keep group acceptance. In a blog conversation,
consistency of theme is created and preserved by commenters controlling how far their
comments diverge from the overall group theme. Commenters balance their desire to
express unique, individualistic ideas with the need to conform to the conversational
54
direction of the group. They thus take the conversation in the direction they perceive the
flock is already going. Thus,
H4: Flocking behavior is negatively related to cognitive diversity.
It is proposed that the flocking process is closely related to the threshold behavior
discussed earlier, in that idea seeders are the emergent flock leaders. Their comments
initiate a cascade of other comments whose thematic consistency exhibits the flocking
phenomenon. There is therefore some probability that there will be two classes of
commenter in a blog entry conversation: idea seeders who often lead the flock and those
who primarily follow.
An estimate of the effect of flocking and idea seeding on a blog entry is
determined in a three step process: separate idea seeders from flock followers (hereafter
denoted flockers), associate the classified individuals with the blog entries they
commented on and then estimate the value of the flocking and idea seeding latent factors
from the participation history of the associated commenters.
Idea seeders can be separated from flockers by a discriminant model (equation
3.2) that differentiates between commenters on the basis of diversity of thought
expressed. Commenter type is assessed on the basis of how prolific their comment
activity is (C), the mean cognitive distance ( S ) between their comments across blog
entries (widest for idea seeders), the standard deviation of the cognitive distances (σS)
between their comments across blog entries (largest for idea seeders), the number of new
thematic clusters they started (K) and the standard deviation of the cognitive distances
55
(σK) between their comments and the centroids of the closest thematic clusters (widest
for idea seeders) across blog entries.
(3.2) 0 1 2 3 4 5S KZj C S Kβ β β β σ β β σ= + + + + +
In equation 3.2 the unit of analysis is the individual commenter, Zj is the
discriminant Z score of the discriminant function j, and the rest of the parameters are as
above. It should be noted that the mean cognitive distance between comments and the
centroids of the closest thematic clusters across blog entries is not expected to
differentiate between idea seeders and flockers because, as described in Chapter II,
flockers can have wide variety in their preferred distance from the rest of the flock. The
distinguishing factor is that there will be a high level of consistency in this distance. That
consistency is fully captured in the standard deviation of the cognitive distances between
their comments and the centroids of the closest thematic clusters across blog entries.
Once commenters are separated into idea seeders and flockers, then the influence
that each type exerts on the cognitive diversity of a blog entry can be modeled as shown
in Figure 3.4. In Figure 3.4, all the parameters of equation 3.2 are useful in assessing the
relationship between idea seeding and cognitive diversity. The relationship between
flocking and cognitive diversity is fully captured by using three indicators: the mean
number of clusters started, the standard deviation of individual comment-to-cluster
centroid distances and the mean number of comments written by commenters to the
current blog entry. This is because flockers are only characterized by being very
consistent in the degree their thought expression over time differs from that of the group
56
and by being more inhibited in their thought expression as a result of being more careful.
Therefore, flockers will comment less and start fewer new thematic clusters. Variation in
the cognitive distance between their comments across blog entries is not important
because that is completely dependent on the thematic perspective of the flock they
belonged to when they make comments to the current blog entry.
FIGURE 3.4
Flocking and Idea Seeding Reflective Sub-Models
57
Memetics and Cognitive Diversity
In Chapter II, memetics was described as the study of mind viruses (i.e., memes).
It was mentioned that Dawkins (2006) described two strategies that evolved to maximize
the success of meme propagation: intense indoctrination and mass dissemination.
Indoctrination uses repetition of the same meme or system of related memes (a
memeset), often while simultaneously excluding or ridiculing competing memes, in a
closed cognitive context (in this case, a blog) to overcome resistance, increase
acceptance and spread of the meme ideas. Mass dissemination spreads a meme or
memeset in as many cognitive contexts as possible, looking for a sympathetic audience
of potential converts with little resistance to the meme ideas. Note that, in the
blogosphere, it is not suggested that mass dissemination occurs under the organizing
influence of some central control. The personality and enthusiasm of individuals drive
them to spread the memes they have embraced in a wide variety of cognitive contexts.
This is proposed, in addition to flocking, to be another example of emergent behavior,
individuals behaving in the same manner, under their own volition, who inadvertently
create the appearance of collective order.
58
FIGURE 3.5
Spectrum of Blog Author Thought Diversity
Alternatively, as noted in Chapter II, indoctrination, as a top-down influence, can
only occur under the influence of a central control. In the blogosphere, that control can
only be a blog author, whose cognitive distance and, probably, time interval between
blog entries are less than average across the blogosphere. Since indoctrination is, by
definition, an extreme behavior it is proposed that blog authors occupy a spectrum of
thought diversity, and by implication, tolerance of diverse thought, as depicted in Figure
3.5. Throughout this discussion of indoctrination, it was assumed that there was some
intent on the part of the blog author to propound a limited set of views. It is possible
though that the appearance of indoctrination might be created by a blog author who
watches the blog to see what topics garner the most discussion and then repeatedly
introduces those topics with no intent but to keep the conversation lively.
59
In Chapter II, indoctrination and mass dissemination were both connected to
Granovetter’s (1978) thresholding dynamic. It was reasoned that indoctrination seeks to
overcome thresholds of resistance to meme ideas through repetition in a closed context,
while mass dissemination seeks to find people with low thresholds of resistance in a
variety of contexts (e.g., those characterized by both cultural tribalism and need-for-
cognition). Just as rioting behavior can spread through a crowd as differing individual
thresholds of riot resistance are overcome by observing others’ rioting behavior, so
meme acceptance can be expected to spread through any group as the unconverted see
others in their vicinity become converted. This cascading propagation of meme
acceptance is expected to be aided by flocking behavior. People with flocking propensity
will sense if their local group’s thematic direction has been affected by the acceptance of
a meme or memeset. In order to preserve group acceptance, they will control the extent
to which their expressed ideas diverge from their perception of the group norm, thus
aiding in memetic propagation.
60
FIGURE 3.6
Locus of Influence Conceptual Map
On the basis of the foregoing, it is further argued that indoctrination and mass
dissemination can be opposing means of memetic propagation. An individual attempting
to disseminate a competing meme in an indoctrination context will face the opposition of
an entrenched set of memes, and a process of constant reinforcement of those existing
memes, that will work to suppress the propagation of a new meme. If the new meme
does manage to take hold and propagate, the commenters who embrace the new meme
will probably migrate, as a cultural tribe, to a different blog context that may indeed be
an indoctrination arena for the new meme. Recall in Chapter II, that indoctrination was
61
denoted as a top-down influence while cultural tribalism was denoted a “grassroots”
phenomenon. It was noted that for indoctrination to have any effect it must garner
grassroots support. It is argued that indoctrination will more often gain its grassroots
support from cultural tribalism, while blog author free thinking will most often gain its
support from need-for-cognition. These proposed relationships are depicted as a
conceptual map in Figure 3.6.
FIGURE 3.7
Evangelists and Idea Seeders
As already stated, mass dissemination is manifested as an individual, rather than
a systemic, activity where certain prolific commenters have low diversity in thought
expression as they repeat the same ideas. These commenters are denoted as evangelists.
This description might make evangelists appear similar to idea seeders; however, idea
seeders are not only prolific but diverse in the expression of their thoughts. The
62
distinction is made more explicit in Figure 3.7 which shows an assumed normal
distribution in the diversity of expressed thought among prolific commenters;
commenters in the low diversity tail are evangelists while those in the high diversity tail
are idea seeders.
FIGURE 3.8
Perceptual Map of Commenter Types
The parameters of equation 3.2 can be re-estimated to distinguish between
evangelists, idea seeders and flockers. To conceptually distinguish between flocking,
idea seeding and evangelism, the parameters of equation 3.2 can be divided into two
63
groups to create sub-factors that represent individual thought diversity (mean and
standard deviation in individual thought separation) and boldness (number of thematic
clusters started, standard deviation in comment-to-closest-thematic-cluster separation
and the number of comments contributed). These sub-factors can be used as axes of the
perceptual map shown in Figure 3.8. If commenters, based on their participation history,
are plotted on this perceptual map, their type should be readily distinguishable by their
presence in one of the marked quadrants.
Merriam-Webster (2007) defines boldness as “showing a fearless, daring spirit.”
Roget (1962) defines it as “willingness to take risks,” (p. 891) with synonyms such as
“adventuresomeness” and “venturesomeness.” These terms and definitions are very
much in keeping with the concept of entrepreneurship, defined by Merriam-Webster as
“assuming the risks of a business.” Another similar term is innovativeness, defined by
Roget as: “being characterized by, or productive of, new things or ideas.” (p. 139)
Clearly, boldness, innovativeness and entrepreneurial are related terms and concepts.
However, entrepreneurial has an explicit business connotation, where boldness is a more
generic term. This difference is important to this study, as a word denoting a willingness
to take risks in expressing ideas is called for, despite the marketing context.
Innovativeness also does not fully capture the need as it involves producing new
ideas, while idea seeders (those high in the quality for which a label is sought) need not
be idea originators; they may also be conduits for ideas acquired elsewhere that they
think are applicable to a particular blog conversation. Idea seeders thus enable what Pink
(2005) calls symphony: taking little pieces of information, and weaving them together so
64
a broader pattern can be discerned (p. 125). Idea seeders, and later the bolder and more
intellectually diverse flockers, individually bring little pieces of information, both
original and from diverse sources, while the collective conversation weaves them
together to find the broader pattern. This makes idea seeders a new type of influential
(Katz and Lazarsfeld 1955) distinct from the concept of mavens and opinion leaders (as
presented by Vernette 2004). The latter were seen as static in their information
dissemination role (Watts and Dodds 2007), while these are information conduits on a
more happenstance basis, dependent on what the conversation calls to mind and what
their boldness allows them to express to the group.
Now that the relevant constructs are defined, the influence of indoctrination and
mass dissemination on cognitive diversity can then be reflectively modeled as shown in
Figure 3.9.
65
FIGURE 3.9
Memetic Propagation Reflective Sub-Models
While it should be apparent from the preceding discussion that cognitive
diversity will be lower in blogs highly influenced by indoctrination, it might be thought
that mass dissemination is just an isolated act of individual fanatics that can have no
effect on overall thought diversity. However, it has already been noted that mass
dissemination can cause memes to enjoy a threshold-overwhelming cascading
propagation of acceptance, aided by flocking behavior. Therefore, both mass
dissemination and indoctrination can be major influences on cognitive diversity. Thus,
H5: Mass dissemination is negatively related to cognitive diversity.
H6: Indoctrination is negatively related to cognitive diversity.
66
Reciprocity and Cognitive Diversity
In Chapter II, it was mentioned that in a blog, the author posts (a costly
effort) in anticipation of response (a payoff). Equation 3.3, a variation of a model
proposed by Gintis (2004), describes the value received by the author ( Av ), the
difference between the author’s valuation of the community’s contribution ( TAB )
and his own cost ( Ac ).
(3.3) A
K
j j
jj
ATAA cb
cBv −−
−=−= ∑
=1
))1(
)1((
δ
ε
Where: K is the number of responders, b is the benefit the responder intended to
contribute (0 < b < 1), ε is the responder’s ineffectiveness in contributing the intended
benefit ( 10 << ε ), δ is the author’s premium (or discount) for the value of each
responder’s contribution (-1< 1<δ ), and Ac is the author’s cost of posting.
Both active (those contributing to the discussion) and passive community
members receive the value from the author’s post and from the contributions of active
members. Equation 3.4 shows the value received by each contributing member ( Cv ) and
equation 3.5 shows the value received by passive members ( Pv ).
(3.4) C
A
AAK
cjj j
jj
CACOCC cbb
cbBv −−
−+
−
−=−+= ∑
≠= )1(
)1()
)1(
)1((
1 δ
ε
δ
ε
(3.5) )1(
)1()
)1(
)1((
1 A
AAK
j j
jj
APTPP
bbbBv
δ
ε
δ
ε
−
−+
−
−=+= ∑
=
67
Where: OCB or TPB is the prospective contributor’s valuation of the contributions of
other active members, ACb or APb is the prospective contributor’s valuation of the
author’s post and Cc is the prospective contributor’s cost of participation.
Equations 3.4 and 3.5 can be applied to the community as a whole. Equation 3.6
is the total value received by a community with at least one participating member ( TCv ).
Equation 3.7 is the value received by a passive community ( TPv ).
(3.6) )1(
)1()
)1(
)1((
1 A
AAK
j
j
j
jj
ATTCTCTC
bc
bbCBv
δ
ε
δ
ε
−
−+−
−
−=+−= ∑
=
(3.7) )1(
)1(
A
AA
ATTP
bbv
δ
ε
−
−==
Where: TCTC CB − is the net benefit of community contribution and ATb is the
community’s valuation of the author’s post.
As a game, this scenario can be described by the following matrix:
Author
Post No Post Reply
ATC vv , Community
No Reply )(, ATP cv − 0,0
If the author posts and the community reciprocates, the community receives certain
payoff TCv while the author receives contingent payoff Av . If the author posts and the
community fails to respond, the community receives certain payoff TPv , its valuation of
68
the author’s post. The community always knows the current value of the discussion and
will only post when the value of existing content exceeds the lowest reciprocation cost
of any member under the terms described below. The author never knows his payout
prior to his post; however, for the author to keep posting in repeated games, he must
expect his payoff to at least exceed his cost ( AA cv > ).
The community maximizes its self-interest differently based on its expectations
of whether the author will keep on posting content of at least minimal quality from its
own perspective (repeated game) or abandon the blog if no or too little response is made
in the first game (non-repeated game). Since the community can always count on the TDv
payout, the TCv payouts are easily calculated. Equilibrium values for the community
payoffs are as given in equations 3.8 and 3.9:
(3.8)
×≤
,2
,
TC
TC
TPC
Cv
repeatednon
repeated
−
(3.9)
×>
,2
,
TC
TC
TCC
Cv
repeatednon
repeated
−
The return to the reciprocating community is greater than that of the passive community
if the game is repeated more than once ( TCC×3 versus TCC×2 ). So, this model
supports the prediction that the community will select repeated game play from the
beginning.
69
Since blogs can be dedicated to a variety of subjects, there can be no fixed
standard by which to assess value, content is valuable if the blog community sees it as
being valuable. This study proposes to objectively measure value under these terms by
adopting a variation of Google’s PageRank algorithm similar to that specified in Dwyer
(2007). Dwyer (2007) observed that Google faced a similar problem as the one faced
here when they wanted to find the most valued web content associated with any search
term. Since webpages can contain content on any topic, and they needed a way of
assessing value without analyzing the content, they adopted a populist criterion for
assessing value whereby the most link-referenced webpages were considered most
valued. According to Bianchini, Gori, and Scarselli (2005), the PageRank ( px ) of page p
is computed by taking into account the set of pages (pa[p]) pointing to p:
(3.10) [ ]
)1( dh
xdx
ppaq q
q
p −+= ∑∈
Where )1,0(∈d d is a proportioning factor and qh is the outdegree of q, that is, the
number of links coming out from page q. The proportioning factor determines the
amount of importance added to p by the pages linking to it. Page p has an inherent
importance of d−1 .
In this study, a blog entry and its comments are modeled as being connected
based on their resonance or similarity of theme. It is proposed that equation 3.10 be
modified to assess the value of a comment as described by equation 3.11:
70
(3.11) [ ]
(1 )p q
q cm p
v d R v d∈
= + −∑ i
In equation 3.11, vp is the value assigned comment p, q denotes all the other comments
with a resonance (R) to p greater than 0. The total value of an author’s entry can be
calculated the same way, as the recursive sum of the values of all resonant comments.
In Chapter II, it was argued that some proportion of readers become commenters
under the influence of a sense of obligation over prior value received from the blog. It is
argued that the total value calculated in equation 3.11 reflects the degree of future
reciprocity that will result from an author’s entry. It is also argued that since some
proportion of commenters will be influenced by reciprocity, the amount of reciprocity
will be proportional to the total number of commenters. These two indicators are
depicted in their relationship to reciprocity in Figure 3.10. Since reciprocity should bring
in commenters from the full spectrum of commenter types, it is argued that:
H7: Reciprocity is positively related to cognitive diversity.
71
FIGURE 3.10
Naïve Model: Hypothesized Relationships
All eight latent variables and hypotheses are integrated into a single conceptual
model presented in Figure 3.10. Definitions of the latent variable indicators are
aggregated in Appendix B.
Alternative Model
In this chapter, the relationships between six social forms and cognitive diversity
have been combined into the naïve model of Figure 3.10. As a naïve model, the social
forms are seen exclusively as acting directly on cognitive diversity. However, since this
research study takes a social process theory perspective, it must be recognized that the
six social forms may be interlinked in a complex manner. While it may seem that these
interlinked social forms are interaction effects that should be captured by the
72
multiplicative interaction terms commonly found in regression models, it is not certain
that the nature of the interlinking will be detected in this manner. The social forms in this
study are often highly dependent on initial conditions, are path dependent, nonlinear and
only loosely coupled to each other. They often also involve random delays and different
timescales. Quantification of the relationships between such processes is seldom
attempted in marketing research and is an issue this study lays a foundation for future
research to address.
Nevertheless, certain interrelationships that can be captured in a conventional
structural equation model have been implied in this discussion. Those interrelationships
have been combined into the model shown in Figure 3.11. In addition to including
hypotheses 1 and 2 from the naïve model, as R8 and R9 respectively, this model depicts
three exogenous influences:
a) The spectrum of blog author thought diversity, a top-down influence acting from
indoctrination to free thinking.
b) Mass dissemination, a grass-roots influence acting on the following individually
motivated or need-based social processes that subsequently influence cognitive
diversity (R8 and R9 in Figure 3.11):
i. Cultural tribalism, inspired by a need for self-expression.
ii. Flocking, motivated by a need for acceptance.
iii. Need-for-cognition.
c) Idea seeding, also a grass-roots influence, acting on the same social processes as
mass dissemination.
73
FIGURE 3.11
Alternative Model: Relationships
It was mentioned in the discussion of Figure 3.5 that the appearance of
indoctrination might be created by a blog author who watches the blog to see what topics
garner the most discussion and then repeatedly introduces those topics with no intent but
to keep the conversation lively. This scenario is included in the model of Figure 3.11
with the inclusion of prior collaborative value (“Total Value T – 1”) as a formative
influence on indoctrination.
In Chapter II, indoctrination and mass dissemination were both connected to
Granovetter’s (1978) thresholding dynamic. It was reasoned that indoctrination seeks to
overcome thresholds of resistance to meme ideas through repetition in a closed context,
while mass dissemination seeks to find people with low thresholds of resistance in a
variety of contexts (e.g., those characterized by both cultural tribalism and need-for-
74
cognition, relationships R2 and R4 in Figure 3.11). Just as rioting behavior can spread
through a crowd as differing individual thresholds of riot resistance are overcome by
observing others’ rioting behavior, so meme acceptance can be expected to spread
through any group as the unconverted see others in their vicinity become converted. This
cascading propagation of meme acceptance is expected to be aided by flocking behavior.
People with flocking propensity will sense if their local group’s thematic direction has
been affected by the acceptance of a meme or memeset. In order to preserve group
acceptance, they will control the extent to which their expressed ideas diverge from their
perception of the group norm, thus aiding in memetic propagation (R5 and R6 in Figure
3.11).
The logic underlying most of the relationships in this alternative model should
now be clear. However, the manner in which flocking and mass dissemination are
depicted may raise some questions. In this narrative, flocking commenters have been
differentiated from evangelists and idea seeders by their lack of boldness. As flockers
are thus characterizable as followers rather than leaders, the presence of flocking is
portrayed in Figure 3.11 as inspired by the more aggressive influences of cultural
tribalism (R5) and need-for-cognition (R6). It is argued that these more aggressive
influences result in the formation of core groups of interaction around whom flockers
gather depending on which group they desire to be affiliated with.
Mass dissemination, as implemented by evangelists, has been depicted in Figure
3.11 as related to both cultural tribalism (R2) and need-for-cognition (R4), while idea
seeding is exclusively related to the latter (R7). To the blog reader, evangelists and idea
75
seeders initially look the same. They are both commenters that often introduce radical
ideas into the blog conversation; only with close study does it become apparent that
evangelists repeatedly propound the same ideas. As a result, both will activate need-for-
cognition in the people to whom the ideas are new. The people who like the evangelized
ideas will naturally form a tribe around them.
Relevance of Sociological Theory
Hulberg (2006) examined the standard practices of corporate image management
using Burrell and Morgan’s (1979) sociological framework. Hulberg notes that Burrell
and Morgan’s framework describes four paradigms, or perspectives, that purports to
classify all sociological theories: (1) Functionalist: Consumers’ image perceptions can
be influenced by corporate messages and behavior; (2) Interpretive: Consumers’ image
perceptions are entirely formed through social interactions among them and cannot be
influenced by corporate messages or behavior; (3) Radical humanism: Consumers view
corporate attempts to convey image as manipulative; and (4) Radical structuralism:
Consumers view all corporate action as motivated toward gaining power and profit that
will be used to suppress others.
Clearly, the image management process of Figure 2.1 is predominantly based on
the functionalist paradigm: companies are seen as able to influence image perceptions.
However, the inclusion of the social forms (socio-psychological processes) hypothesized
to be the primary mediators in the expression of consumers’ thoughts to blogs, allows
the use of not only functionalist, but interpretive and radical humanist perspectives:
76
a) Functionalist. It is readily apparent that the memetic processes of indoctrination
and mass dissemination explicitly assume that one party can influence the image
perceptions of another. Even though idea seeding and reciprocity do not result in
perceptions of image being influenced, they do initiate cascades of behavior
where individuals influence one another by causing boldness and sense of
obligation thresholds to be exceeded. Idea seeding and reciprocity can thus be
viewed from a functionalist perspective.
b) Interpretive. Flocking and cultural tribalism are both processes where outcomes
depend on group norms established without the intervention of the blog author or
company hosting the blog. The degree of cognitive diversity expressed by
flockers will be dependent on individual perceptions of how much divergent
expression the group will tolerate. Since cultural tribalists are motivated by a
need for self expression, they will be attracted to blog communities already
focusing on their preferred themes. The tribe can be expected to migrate to any
blog whose theme more closely matches their preference, abandoning any blog
whose theme changes sufficiently. Therefore, even though cultural tribalism is
supported by indoctrination that matches tribal preference, the phenomenon is
caused by like individuals coming together and thereafter acting in concert.
c) Radical humanist. As discussed in the section on memetic processes,
indoctrination and mass dissemination are specifically intended to be tools of
persuasion. Thus, any blog readers who are motivated primarily by need-for-
cognition are expected to recognize these influences for what they are and will,
77
depending on the intensity of their use, either disregard them or abandon the blog
for being too narrow in subject matter. Blog readers motivated by the need for
self expression underlying cultural tribalism are also expected to have a similar
response. Note that blog readers whose preferences match ideas spread through
these memetic influences will not see them as manipulative.
While the association of sociological processes with image management is not
new, this study increases awareness of the variety of ways that sociological processes
influence image diffusion and the expression of image.
Summary
Earlier in the chapter, the focus of this study was stated to be investigating the
effects of six social forms, identified by prior research to strongly affect cognitive
diversity, the diversity of thought expressed in a group forum. If the operation of these
social forms cause the expression of consumer brand image to be distorted or
incomplete, then companies that mine blogs for brand image insights will not gain an
accurate picture of brand image in the market. To that end, measurable constructs for the
six social forms have been proposed as eight latent variables with suitable indicators.
These latent variable constructs and their indicators are summarized in Table 3.1. These
latent variables have been configured into two models, a naïve model proposing direct
effects on cognitive diversity and an interaction model proposing interrelationships
between seven of the latent variables. Full depictions of these models, with all reflective
and formative relationships, are included in Appendix C as Figures C-2 (Naïve) and C-3
78
(Alternative). The next chapter details the methodology used to test the hypothesized
relationships.
79
TABLE 3.1
Hypotheses and Construct Definition Summary
Constructs Indicators Mean cluster density. After the comment clusters for each blog entry are
found, the distance from each comment’s location in cognitive space to the
centroid of its cluster can be calculated.
Mean cluster radius. After the comment clusters for each blog entry are
found, the distance from each comment’s location in cognitive space to the
centroid of its cluster can be calculated.
Number of clusters. After the comment clusters for each blog entry are
found, the number of clusters is proposed to be indicative of the number of
conversational themes.
Cognitive Diversity. The many
cognitive styles or ways
individuals think, perceive and
remember information and then
use that information to solve
problems or, in general, enact their
behavior.
Percentage of outliers. After the comment clusters for each blog entry are
found, the distance from each comment’s location in cognitive space to the
centroid of its cluster can be calculated. The mean and standard deviation of
each cluster’s comment-to-centroid distances is calculated and a z-score
calculated for each comment. Any comment with a z-score greater than 2.0
is considered an outlier.
Mean collective thought separation. The mean of entry-to-comment and
comment-to-comment cognitive distances for the current blog entry.
Stdev collective thought separation. The standard deviation of entry-to-
comment and comment-to-comment cognitive distances for the current blog
entry.
H1 Cultural Tribalism. A state that a
virtual community can be in where
the members share a strong
ideological identity, creating a
kind of intellectual segregation
where like only talks to like.
Mean indiv. thought separation. The mean comment-to-comment cognitive
distance, across blog entries, between comments made by the same
individual, a commenter on the current blog entry.
80
TABLE 3.1 (CONT)
Constructs Indicators Stdev indiv. thought separation. The standard deviation of entry-to-
comment and comment-to-comment cognitive distances for the current blog
entry.
Mean commenter longevity. The mean time between the first comment and
the current comment of all commenters to the current blog entry.
H1
(cont)
% Repeat commenters (T - 1). The number of commenters contributing to
the current blog entry who contributed to the previous blog entry.
Mean collective thought separation. The mean entry-to-comment and
comment-to-comment cognitive distances for the current blog entry.
Stdev collective thought separation. See above.
Mean indiv. thought separation. The mean entry-to-comment and comment-
to-comment cognitive distances for the current blog entry.
Stdev indiv. thought separation. See above.
H2 Need-for-Cognition. Part of what
determines the level of thought
exercised is innate need-for-
cognition, an enjoyment of
cognitively demanding tasks
(From Petty and Cacioppo’s
(1986) Elaboration Likelihood
Model).
% First time commenters. The number of commenters contributing to the
current blog entry who have never contributed to previous entries.
Mean clusters started. The mean number of thematic clusters started over
the participation lifetime of a blog commenter.
Stdev indiv. thought separation from cluster centroid. The standard
deviation of the cognitive distance between a commenter’s comments and
the centroids of the closest thematic cluster across blog entries over the
participation lifetime of a blog commenter.
Mean number of comments. The mean number of comments (across blog
entries) created by individuals commenting on the current blog entry.
Mean indiv. thought separation. See above.
H3 Idea Seeding. The most
uninhibited commenters
expressing ideas on the most
diverse array of themes.
Stdev indiv. thought separation. See above.
81
TABLE 3.1 (CONT)
Constructs Indicators Mean clusters started. See above.
Stdev indiv. thought separation from cluster centroid. The mean cognitive
distance between a commenter’s comments and the centroids of the closest
thematic cluster across blog entries over the participation lifetime of a blog
commenter.
H4 Flocking. People control the
expression of their thoughts to
balance individuality with group
cohesion.
Mean number of comments. See above.
Mean clusters started. See above.
Stdev indiv. thought separation from cluster centroid. See above.
Mean number of comments. See above.
Mean indiv. thought separation. See above.
H5 Mass Dissemination. One of the
two primary memetic propagation
mechanisms where the targets are
the people with the lowest barriers
to adoption, so the idea is spread
as widely as possible to increase
the probability of reaching people
with a low threshold of resistance.
Stdev indiv. thought separation. See above.
Mean time between entries. The mean interval between successive entries
posted to a blog.
Mean entry-to-entry separation. The mean pair-wise cognitive distances
between entries to the same blog.
H6 Indoctrination. One of the two
primary memetic propagation
mechanisms where the target
person or population is subjected
to a high repetition of the meme
idea in an attempt to overwhelm
thresholds of resistance.
Stdev entry-to-entry separation. The standard deviation in the pair-wise
cognitive distances between entries to the same blog.
H7 Reciprocity. As people read the
blog and get value from it, a sense
of obligation to contribute builds
up.
Total number of commenters. The total number of unique commenters to a
blog entry.
82
CHAPTER IV
METHODOLOGY
Chapter III introduced an integrated naïve model of how six social processes
interact to affect cognitive diversity and an alternative model portraying certain
interrelationships between the six social processes. In this chapter, a means of testing the
explanatory sufficiency of these models is described. It will be described how the model
in Figure 3.10 was instantiated as a multi-group model. It will be shown how blog entry
comments were separated into those influenced primarily by cultural tribalism and those
characterized by need-for-cognition in so much as they embody the attributes of the
presence of those grassroots influences. Within each blog, commenters were segmented
into idea seeders, evangelists and flockers as they embodied the influences of threshold
behavior, mass dissemination and flocking, respectively. All commenters, regardless of
type were predicted to be capable of engaging in behavior motivated by reciprocation,
need-for-cognition and cultural tribalism. The aggregate model simultaneously captured
the influence of the full spectrum of blog author thought diversity levels and the three
groups of commenters on overall cognitive diversity.
This chapter begins by describing the data set used in this study and the
considerations made in its selection. Then, a step-by-step description is given of how the
raw data was processed to make it suitable for testing the hypotheses presented in
Chapter III. Next, the criterion used in determining whether or not hypotheses are
83
supported is described. Finally, the limitations associated with the methodology are
described.
Data
Kozinets (2002) introduced netnography as a methodology where the principles
of ethnography, or unobtrusive observation, were applied to the study of a virtual
community of coffee aficionados. He specified a selection criterion for subject
communities that differed from the practice of standard ethnography: select communities
that are focused on a research question-relevant topic, receive above average posting
traffic, have a large number of contributing members, contain descriptively rich content
and thus, in general, enjoy a high level of member-to-member interaction. Having a high
level of member-to-member interactions is important to this study because it increases
the number of cognitive objects (i.e., bodies of text) and, as a result, the probability of
detecting the effects of the social processes described in Chapter III.
The study used author posts and community response data from the 15 diverse
blogs listed in Table 4.1. Consistent with the netnography methodology proposed by
Kozinets (2002), these blogs were purposively selected because they are:
a) Highly active. Blogs whose authors post frequently and each post tends to attract
more than 20 comments from readers (Table 4.2).
b) Representative of the wide diversity in blog topics relevant to marketing. In light
of the broad scope of marketing as a field of study, as summarized in Table 4.1,
in addition to blogs related to companies and products, blogs related to political
84
ideology, popular culture and people who are business thought-leaders are
included in the study.
c) Relevant to the research questions. Indoctrination is a social process whose
effects are not likely to be strong in all blogs. Therefore, three blogs (Blog for
America, The Evangelical Outpost and Townhall) were purposely selected as
indoctrination blogs because an informal examination of their posts and
comments indicated a high level of narrowly-scoped ideological content. Three
additional blogs (EnGadget, Gizmodo and Joystiq) were selected as need-for-
cognition blogs because they frequently introduce new products to consumers.
The overview of need-for-cognition and sensemaking theory presented in
Chapters II and III suggests that such blogs might prompt a higher level of
cognitive activity motivated by need-for-cognition as consumers collaboratively
converge on an understanding of the benefits that these new products might offer.
The remaining blogs were selected because their high level of activity should
allow most of the processes studied to be manifested.
d) Allow comments from readers. In Chapters I and II, it was emphasized that the
purpose of blogging was for consumers and producers to converse.
Unfortunately, many corporate blogs do not allow for reader comments. Since
this study focuses on the expression of diverse thought, only blogs that allow
comments are suitable.
85
e) Retain an archive of posts and comments. Some blogs only show the most recent
activity. Hypotheses 1, 2 and 4 require that the long term activity of commenters
be studied to detect regularities in the commenting activity of individuals.
TABLE 4.1
Data Sources
3 The number within parentheses is the weblog’s position on the Technorati, a weblog search engine,
ranking of the most authoritative (most widely cited by other websites) of the almost 100 million weblogs
tracked (as of Aug 07). An asterisk denotes situations where the whole blog archive was used because
there were fewer than 2000 blog entries.
ID Blog Name Description 1 AutoBlog (71)
3 A blog covering the auto industry with test drives
and commentary on articles from other sites.
2 Blog for America
(21,406)
The Democratic Party’s blog.
3 Blog Maverick (269)* Entrepreneur Mark Cuban’s blog.
4 The Consumerist (24) A consumer retaliation blog.
5 EnGadget (1) A product review blog for high-tech consumer
goods.
6 The Evangelical
Outpost (1668)
Reflections on culture, politics and religion from an
evangelical worldview.
7 Fastlane (8890)* GM Vice Chairman Bob Lutz’s blog.
8 Freakonomics (137)* The blog for the book Freakonomics
9 Gizmodo (3) A product review and news blog for high-tech
consumer goods.
10 Google Blogoscoped
(100)
A consumer’s blog covering Google and its
services.
11 Joystiq (39) A product review blog for the gaming industry.
12 PaulStamatiou (99)* A college student reviewing technology products
and offering technical support
13 Townhall (73)* Hugh Hewitt’s, a radio talk show host, blog on
politics, religion and conservative social
commentary.
14 TV Squad (104) Commentary and review of TV shows.
15 The Unofficial Apple
Weblog (52)
Since Apple has no corporate blog, their fans have
started blogs about them.
86
TABLE 4.2
Data Source Quantities
Blog Dates Posts Comments Commenters AutoBlog Mar 07 – Jul 07 2,036 27,826 14,657
Blog for
America
Mar 03 – Dec 04 2,232 252,395 15,388
Blog Maverick Nov 99 – Sep 07 435 16,681 7,546
The
Consumerist
Mar 07 – Aug
07
2,599 73,052 9,141
EnGadget May 07 – Jul 07 2,030 51,738 14,203
The Evangelical
Outpost
Oct 03 – Sep 07 1,731 52,027 4,813
Fastlane Jan 05- Aug 07 266 15,914 5,572
Freakonomics Mar 05 – Aug
07
1,142 23,445 8,375
Gizmodo Jul 07 – Aug 07 2,243 39,091 7,078
Blogoscoped
Jan 06 – Aug 07 2,460 20,227 4,015
Joystiq Jun 07 – Sep 07 2,058 52,130 7,218
Paul Stamatiou Sep 05 – Sep 07 757 10,195 3,102
Townhall Apr 06 – Sep 07 985 15,455 2,689
TV Squad Jan 07 – Jul 07 2,034 24,994 6,605
The Unofficial
Apple Weblog
Apr 07 – Sep 07 2,165 23,414 7,465
Total 25,173 698,584 117,867
Two of the blogs (Blog for America and GM Fastlane) were selected as pretest
blogs because they are expected to demonstrate activity at the extremities of the various
conceptual maps (i.e., Figures 3.2 and 3.6) used to demarcate the diverse influences
affecting blogs.
87
Method
Data Collection and Preprocessing
The first phase of this study was to collect a sample of approximately 2000
authors’ posts and associated comments from each blog using a custom software
program that automated the collection. Automated processes, used throughout this study
and noted where appropriate, are particularly relevant to this study because of the large
volume of data. In cases where the blog had less than 2000 author posts, the entire blog
archive was collected. Then, each entry and comment was decomposed into centering
resonance analysis (CRA) word networks, the influence of each node of the word
network was calculated, and then the resonance between posts and comments was
calculated on the following dimensions:
a) Post-to-post. The resonance of a post with all posts that follow it. This was
designed to show the extent to which all posts follow the same theme.
b) Post-to-comment. The resonance of a post with the comments that respond to it
and the resonance between the responding comments. This was designed to
indicate the extent to which the comments thematically diverge from that of the
post.
c) Prior post and its comments to subsequent posts. The resonance between past
posts and their comments with future posts. This was designed to reveal the
extent to which authors are responsive to issues important to the commenters.
88
d) Posts by the same author. Since some blogs have more than one author this
dimension was designed to indicate whether individual authors keep posting on
the same theme.
e) Comments by the same commenter. Capturing the diversity of thought expressed
by individual commenters was an important part of assessing the degree of
autonomy tolerated by the group.
f) Comment-to-comment. The resonance between all comments, regardless of
associated post, was designed to allow clusters to be detected that indicate major
themes or the content of the community’s common knowledge structure.
Corman and Dooley (2008), two of the authors of Corman et al. (2002), and
inventors of Centering Resonance Analysis (CRA), founded Crawdad Technologies,
LLC to market Crawdad, a software program that automates the process of turning text
into CRA word networks, and calculating influence and resonance. They permit the
download of an evaluation copy of their software that processes a body of example text
into a CRA word network and performs the relevant calculations. After reading Corman
et al. (2002) and Wasserman and Faust (1994) it was possible to write a program in
Microsoft’s C-Sharp (C#) programming language that duplicated the output of the
Crawdad software using its example text. The C# program was then used to process the
data for this study.
MDS and Cluster Detection
As described in Chapter II, since CRA’s resonance metric allows a pair-wise
measure of similarity to be calculated for every document (i.e., post and comment);
89
Kamada and Kawai’s (1989) spring-like methodology was adapted to self-assemble
clusters of cognitive objects in a multidimensional space. These clusters were then
demarcated using Chiu et al’s (2001) two-step methodology and cluster membership
confirmed with McQueen’s (1967) k-means.
Spring model multidimensional scaling was implemented with a custom program
written in C# that utilized a modified version of Dwyer’s (2004) WilmaScope 3D graph
visualization library. WilmaScope was modified to use four dimensions instead of three
so as to allow the spring models sufficient freedom to assume their desired unstressed
lengths as explained in Chapter II in the sections entitled Multidimensional Scaling and
Spring Models in Multidimensional Scaling.
Negative and Positive Valance
CRA focuses on the detection of thematic consistency between documents;
however, it does not capture differences in positive and negative valence in its
determinacy of theme consistency. To address this shortcoming, a separate accounting of
relative frequency in the use of negative words between two compared documents was
maintained. In the previous section, a means of locating a document in a
multidimensional thematic space is described where relative differences are turned into
relative coordinates that can then be used in the detection of thematic clusters. The
accounting of negative valance is used as an additional dimension in the detection of
thematic clusters.
90
Multigroup Analysis
After the comment text was transformed into cognitive objects located in clusters
in multidimensional space, the task of identifying blog entry comment (cultural tribalism
and need-for-cognition) and commenter (evangelist, idea seeder or flocker) types began.
This study regards the commenters in the blogs listed in Table 4.1 as cluster samples
from the same larger population, people who comment in blogs. Cluster samples differ
from random samples in that subjects in the same cluster are more similar to each other
than those in a truly random sample. Recognizing a level of similarity seems appropriate
because blogs tend to be topical and thus attract people who, at the least, have a common
interest in the same topic. As a result, the parameters for equations 3.1 and 3.2, the
discriminant models identifying primary influences and commenter types, were
estimated separately using commenter data from each of the 15 blogs. Then, the
parameter estimations were compared to check for consistency. Additionally, it was
necessary to demonstrate both the factorial invariance of the measuring instrument (i.e.,
the 8 factors proposed in Chapter III) and the invariance of the causal structure (the
models in Figures 3.10 and 3.11) across the individual blogs. EQS 6.1 was used in
conjunction with the procedures prescribed by Byrne (2006), and summarized below, to
demonstrate both types of invariance.
In order to use equations 3.1 and 3.2 for segmentation, the parameter coefficients
must have been estimated using data already classified. Discriminant models are then
estimated using a portion (typically 70%) of the pre-classified dataset and the accuracy
of the determinant model’s ability to classify data items is then tested using the
91
remaining or holdout sample. Since it is desired to determine whether the number of
clusters naturally present in the data matches the conceptual map topologies given in
Figures 3.2 (cultural tribalism and need-for-cognition) and 3.8 (flockers, idea seeders
and evangelists), Chui et al’s (2001) methodology was used to get an initial cluster
demarcation. Then, the assignment of objects to clusters was confirmed with McQueen’s
(1967) widely used k-means clustering algorithm. When the data was thus pre-classified,
the parameters of the discriminant models (equations 3.1 and 3.2) described in this study
were estimated.
Classification of Blog Entry Comments
Chapter III described how blog entry comments can be classified as either
influenced primarily by cultural tribalism or need-for-cognition using the discriminant
model equation 3.1 or by plotting the positions of comments in the conceptual map
shown in Figure 3.2, derived from equation 3.1. The two comment types were
distinguished, one from the other, using Chui et al’s (2001) methodology using all the
comments from a random sample of 25 blog entries selected from each of the two pretest
blogs. Two clusters emerged and cluster membership was confirmed using McQueen’s
(1967) k-means. When this pretest sample was plotted on the conceptual map of Figure
3.2, the result is as displayed in Figure 4.1.
92
FIGURE 4.1
Pretest Blog Entry Comment Mapping
While the results are not perfectly conforming to the proposed perceptual map
regions, the higher density of data points within the so-called cultural tribalism and
need-for-cognition quadrants suggests support for the proposed topology for blog entry
93
comment segmentation. A notable aspect of these results is that a near equal number of
the comments within the cultural tribalism quadrant belong to the GM Fastlane blog, not
the blog predicted to be an indoctrination blog. A positive aspect of this observation is
that the topology-based segmentation seems to distinguish one comment from another;
however, the unexpected inability to distinguish between blogs suggests that the logic
underlying the initial prediction of blog types may have been too superficial.
The parameters of equation 3.1 were estimated from the pretest dataset and the
accuracy of the resulting model tested with a holdout sample. One discriminant function
emerged from the estimation with a characteristic root eigenvalue greater than 1.0. This
function correctly classified 88.3% of the holdout sample.
Classification of Commenters
Chapter III described how commenters can be characterized as evangelists, idea
seeders and flockers by plotting their positions on a conceptual map (Figure 3.8, derived
from equation 3.2). Idea seeders can be separated from evangelists by high diversity in
thought expression: idea seeders express thoughts about a wide variety of things while
evangelists always express similar thoughts. Idea seeders and evangelists can be
separated from flockers on the basis of differences in boldness: idea seeders and
evangelists are bolder when expressing their thoughts.
When the commenting activity of 1,000 commenters, 500 randomly selected
from the Blog for America (BFA) and the GM Fastlane blog (GM) as a pretest sample,
was demarcated by Chiu et al’s (2001) methodology, three clusters emerged (Figures 4.2
94
and 4.3). Cluster assignments were confirmed using McQueen’s (1967) k-means
methodology.
FIGURE 4.2
Pretest BFA Commenter Clusters
95
FIGURE 4.3
Pretest GM Fastlane Commenter Clusters
The placement of these clusters was considered conclusive and conforming to
expected patterns. It is interesting to note that idea seeding seems to be more pronounced
in the GM Fastlane blog.
96
In preparation for estimating the parameters of equation 3.2 the commenters
were forcibly classified into the segments of Figure 3.8. Using these topology-based
segment membership assignments, the parameters of equation 3.2 were estimated and
the accuracy of the resulting model tested with a holdout sample. Two discriminant
functions emerged from estimation with each of the two pretest datasets, correctly
classifying 86.5% (GM) and 79.3% (BFA) of the holdout sample.
Construct Validity
Bagozzi, Yi and Phillips (1991) broadly define construct validity “as the extent to
which an operationalization [(i.e., a measuring device)] measures the concept it is
supposed to measure.” (p. 421) Construct validity focuses on the extent to which data
exhibits (1) convergent validity, the extent to which different assessment methods
concur in their measurement of the same construct (also called factorial or measurement
invariance), and (2) discriminant validity, the extent to which the measures of different
constructs are distinct (Campbell and Fiske 1959). The demonstration of factorial
invariance began by establishing baseline confirmatory factor analysis (CFA) results for
each blog dataset. The results were then compared to see if the same indicators loaded
on the same factors. After the factorial structure was found to be the same across blogs,
hypotheses that each free factor loading (each factor’s “first” indicator is constrained to
unity and is thus not free) was equal across blogs was tested by forcing the factor
loadings to be equal and then observing overall fit statistics as well as the LaGrange
Multiplier (LM) Test univariate probability for each factor loading. This probability is a
measure of significance for the LM Test, a test of the hypothesis that a factor loading
97
varies across groups (in this case, blogs). LM Test probabilities less than 0.05 are
considered statistically insignificant. That is, when the LM Test probability is less than
0.05, the parameter does not have the same value across the blogs. This procedure does
not require equal sample sizes, nor does the test of causal structure invariance described
below.
Invariance of causal structure was tested in a manner similar to the above. The
same structural model was simultaneously fitted to the 15 blog datasets while
constraining all factor loadings, structural path loadings and error covariances equal
across the blogs. Then multigroup model fit statistics were observed as well as the LM
Test univariate probability for each constrained parameter.
Underlying the use of structural equation modeling to perform CFA is a strong
assumption of multivariate normality. If evidence shows that this assumption is violated,
it has important implications for the interpretability of the findings. None of the
indicators in the pretest data set demonstrated significant nonzero univariate skewness
(one tail of the distribution is longer than the other) or kurtosis (variance is due to
infrequent extreme deviations from the mean). Normalized estimates of Mardia’s (1970)
multivariate coefficient, an aggregate indicator of kurtosis, were found to be 2.15 (BFA)
and 3.98 (GM). Bentler (2005) suggests that only values greater than 5.0 indicate non-
normally distributed data. Another test of multivariate normality is discussed later.
The individual pretest blog baseline CFA results showed the same indicators
loading on the same factors as given in Table 4.3. Additionally, as Table 4.3 shows, all
but eight of the 32 core factor construct indicators met Hair et al’s (2005) criterion for
98
statistical significance (factor loading > 0.4 with a sample size > 200) and 14 met the
criterion for practical significance (loading > 0.5). It must be noted that Hair et al.’s
criteria, as well as most of the reference values quoted in this study, applies to evaluating
instruments used to gather experimental data. The circumstances of this study are much
different as Hair et al.’s criteria are being used to evaluate a means of gaining insights
from non-experimental, albeit primary, empirical data. Therefore, it is argued that Hair
et al’s criteria must be referenced with caution as the level of control over extraneous
variation that characterizes experiments is absent here. That being understood, the
general conformance of this study’s factor loadings and other metrics to those
recommended by Hair et al. (2005) and Byrne (2006) is worthy of note and indicative of
a reasonable level of performance. The multigroup CFA with factor loadings constrained
equal across blogs had good fit (CFI = 0.954, RMSEA = 0.019, χ2 (896) = 1,225)
indicating that both blogs have similar, if not the same, factor structures.
99
TABLE 4.3
Pretest Confirmatory Factor Analysis, Reliability, Variance Explained and Factorial Invariance
Main Factor and Reliability
Indicator Loading R2 LM Test Probability
Mean Cluster Density -0.562 0.317 0.000
Mean Cluster Radius 0.668 0.448 0.000
Number of Clusters 0.612 0.377 0.000
Cognitive Diversity
r = 0.765
Percentage of Outliers 0.537 0.290 N/A
Mean Collective Thought Separation -0.496 0.246 0.000
Stdev Collective Thought Separation -0.571 0.326 0.000
Mean Indiv. Thought Separation -0.493 0.243 0.000
Stdev Indiv. Thought Separation -0.450 0.211 N/A
Mean Commenter Longevity 0.349 0.122 0.072
Cultural Tribalism
r = 0.683
% Repeat Commenters (T - 1) 0.493 0.244 0.895
Mean Clusters Started -0.562 0.316 0.040
Stdev Indiv. Thought Separation From Cluster
Centroid -0.639 0.408 N/A
Flocking
r = 0.625
Mean Number of Comments -0.618 0.381 0.000
Mean Clusters Started 0.411 0.169 0.092
Stdev Indiv. Thought Separation From Cluster
Centroid 0.419 0.177 0.137
Mean Number of Comments 0.470 0.221 0.039
Mean Indiv. Thought Separation 0.365 0.134 0.746
Idea Seeding
r = 0.519
Stdev Indiv. Thought Separation 0.413 0.172 N/A
100
TABLE 4.3 (CONT)
Main Factor and Reliability
Indicator Loading R2 LM Test Probability
Mean Time Between Entries -0.623 0.406 N/A
Mean Entry-to-Entry Separation -0.580 0.337 0.008
Indoctrination / Free
Thought
r = 0.485 Stdev Entry-to-Entry Separation -0.474 0.225 0.002
Mean Clusters Started 0.435 0.216 0.000
Stdev Indiv. Thought Separation From Cluster
Centroid 0.501 0.256 0.006
Mean Number of Comments 0.506 0.259 0.023
Mean Indiv. Thought Separation -0.515 0.278 0.000
Mass Dissemination
r = 0.663
Stdev Indiv. Thought Separation -0.562 0.316 N/A
Mean Collective Thought Separation 0.478 0.230 0.001
Stdev Collective Thought Separation 0.601 0.361 0.195
Mean Indiv. Thought Separation 0.551 0.304 0.001
Stdev Indiv. Thought Separation 0.490 0.251 N/A
Need for Cognition
r = 0.640
% First Time Commenters 0.391 0.153 0.056
Reciprocity
r = 1.0
Total Number of Commenters
1.0
1.0
N/A
Note: ρ < .05. Reliabilities are calculated from the indicators to each individual factor in Table 4.3. Pearson “r” is reported
instead of Cronbach’s α. N/A in the right hand column denotes the arbitrary first factor.
101
Table 4.3 also shows that 13 of the 32 factor loadings exhibited a less than 5%
LM Test univariate probability, indicating that their factor loadings are not the same
across the two blogs even though the structure appears to be the same.
TABLE 4.4
Pretest Correlation and Φ-matrix of Primary Constructs
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.033 -
3 Flocking 0.004 0.001 -
4 IdeaSeed 0.265 0.084 -0.014 -
5 Indoc 0.079 0.475 0.078 0.166 -
6 MassDiss 0.120 0.020 0.206 0.528 0.574 -
7 NFC 0.117 0.183 0.109 0.162 0.942 0.586 -
8 Recip 0.205 0.108 -0.062 0.117 0.981 0.209 0.439
*Note: Correlations are corrected for attenuation. All values are significant to ρ < .05.
CogDiv = Cognitive Diversity, CultTrib = Cultural Tribalism, IdeaSeed = Idea Seeding,
Indoc = Indoctrination, MassDiss = Mass Dissemination, NFC = Need-for-Cognition,
Recip = Reciprocity.
Discriminant validity was assessed in two stages: (1) correlations between
constructs, corrected for attenuation, were confirmed significantly less than unity (Table
4.4 shows the correlations for the GM Fastlane blog), and (2) a χ2 difference test was
performed between the baseline multigroup CFA model (CFI = 0.998, χ2 (872) = 888)
and an alternative multigroup model (CFI = 0.106, χ2 (928) = 7,338) where the
correlation between each latent construct was constrained at unity. According to Cheung
and Rensvold (2002), a large CFI or χ2 difference demonstrates discriminant validity.
While there is a widely accepted understanding of what constitutes a significant χ2
102
difference for a given degrees of freedom, Cheung and Rensvold’s recommendation that
a ∆CFI > 0.01 be considered significant is only beginning to gain acceptance (Byrne
2006, p. 341). In this study, an attempt is made to disclose both metrics where possible.
It is readily apparent that a ∆CFI of 0.892 and a ∆χ2 of 6,450 with degrees of freedom
equal to 56 argue strongly in favor of discriminant validity.
As has already been intimated, prior to conducting a confirmatory factor analysis,
and the model estimation, it is important to test for multivariate normality in the factor
indicators. This is particularly important in this study as many measures of collective
phenomena have an underlying power-law distribution, as discussed in Chapter II, rather
than a normal distribution. Hair et al. (2005) point out the difficulty in testing
multivariate normality (the joint normality of more than one variable) and suggest that
testing univariate normality is an acceptable substitute, especially if sample sizes are
large. Hair et al. recommend graphical analysis using the normal probability plot as well
as statistical techniques such as skewness and kurtosis. Figures 4.4 and 4.5 show normal
probability plots for estimated values of the BFA and GM Fastlane blog’s cognitive
diversity latent variable. These plots show no normality problems.
103
FIGURE 4.4
Pretest Normal Probability Plot of BFA Cognitive Diversity
104
FIGURE 4.5
Pretest Normal Probability Plot of GM Fastlane Cognitive Diversity
Hypothesis Testing
Blog entry comments were characterized as influenced primarily by cultural
tribalism or need-for-cognition and commenters in each blog were segmented into
evangelists, idea seeders and flockers. Then, the models depicted in Figures 3.10 and
105
3.11 were estimated and hypotheses tested. Although this model is fully described in
Chapter III, the segmentation of blog entry comments and commenters might be a source
of confusion. So, before the estimation of the model is discussed the implementation
details involving the segmented blog entry comments and commenters will be described.
Since the unit of analysis in the models of Figures 3.10 and 3.11 is the blog
entry, values must be calculated for the 32 indicators used to reflect the eight latent
variable constructs for each blog entry. As explained in Chapter III, the need for
validated self-expression (as manifested in cultural tribalism) and need-for-cognition are
competing motivations within blogs. Each blog entry is approached in retrospect,
complete with all its comments. As each blog entry is examined, its comments are
segmented based on equation 3.1 into those influenced primarily by cultural tribalism
and those primarily influenced by need-for-cognition. Then summary measures of each
segment, mean and standard deviations of individual and collective thought separation,
are calculated to measure the effect of each segment, and its associated influence, on the
whole blog entry. These summary measures are the indicators for the cultural tribalism
and need-for-cognition latent variables in Figures 3.10 and 3.11.
The flocking, mass dissemination and idea seeding factors reflect relationships
between summary measures of cognitive objects produced by individual commenters.
For example, when calculating the values for the three indicators used to reflect flocking,
all the data for the commenters previously identified as flockers who commented on the
current blog entry are retrieved. From this data the mean number of clusters started and
the mean number of comments posted, across all blog entries, is calculated.
106
Additionally, for each flocker who commented on the current blog entry, the standard
deviation of the cognitive distance between all the comments they post in the lifetime of
their participation and their associated cluster centroids is calculated. The joint reflective
impact of the three indicators on the flocking latent variable becomes the impact of that
construct on the cognitive diversity of the current blog entry.
The calculation of the indicators reflective of the other four latent variables is
fully described in Chapter III and should be readily understood. As a result, the
discussion will now turn to estimating the model.
TABLE 4.5
Pretest Multigroup Fitted Naïve Model Parameter Values
Parameter Value Hypothesis Support
Standardized Unstandardized Statistical Practical LM Test
Probability H1 -0.0625 -0.0543 (t = -1.64) Yes Yes 0.000
H2 0.6469 0.5435 (t = 9.93) Yes Yes 0.000
H3 -0.0300 -0.0372 (t = -0.68)* No No 0.026
H4 -0.1042 -0.0843 (t = -2.54) Yes Yes 0.000
H5 -0.0017 -0.0017 (t = -0.04)* No No 0.000
H6 -0.0748 -0.0721 (t = -1.61) Yes Yes 0.000
H7 0.0015 0.0007 (t = 0.02)* No No 0.000
*Not significant
Prior to fitting the 32 indicator measurements to the models in Figures 3.10 and
3.11, with a pretest sample of 150 blog entries randomly selected from the GM Fastlane
blog and the Blog for America, the models were translated into simultaneous equations.
Then, multigroup structural equation modeling and maximum likelihood estimation
(using EQS 6.1), with factor loadings and relationships between factors constrained
107
equal across blogs, was used to solve the naïve and alternative models and obtain the
results reported in Tables 4.5 (Naïve: CFI = 0.568, RMSEA = 0.121, χ2 (78) = 647) and
4.6 (Alternative: CFI = 0.883, RMSEA = 0.063, χ2 (78) = 232), and Figures 4.6 and 4.7.
TABLE 4.6
Pretest Multigroup Fitted Alternative Model Parameter Values
Parameter Value Relationship Support Path
Standardized Unstandardized Statistical Practical LM Test
Probability R1 0.3757 0.4354 (t = 6.07) Yes Yes 0.000
R2 -0.2108 -0.2427 (t = -4.34) Yes Yes 0.034
R3 -0.0993 -0.1159 (t = -1.93) Yes Yes 0.000
R4 0.1398 0.1720 (t = 3.00) Yes Yes 0.000
R5 0.3021 0.3258 (t = 6.04) Yes Yes 0.003
R6 -0.4416 -0.4498 (t = -8.13) Yes Yes 0.000
R7 0.2761 0.3993 (t = 4.61) Yes Yes 0.010
R8 -0.0862 -0.0770 (t = -2.24) Yes Yes 0.000
R9 0.6483 0.5468 (t = 10.28) Yes Yes 0.000
As shown in Table 4.5, only H1, H2, H4 and H6 are supported as indicated by
both statistical significance and the size of the unstandardized parameter (i.e., effect
size). However, statistical significance is not noteworthy in the main part of this study,
as the large sample size is assured to raise statistical power enough to make all the
estimated values statistically significant. A more meaningful criterion is whether the
sizes of the estimated parameter values (i.e., effect size) show they have practical
significance (Kirk 1996). Kirk eschews the selection of a static and universal threshold
value of practical significance to be applied to every scenario. This study arbitrarily
defines practical significance as an unstandardized parameter estimate greater than 0.05.
108
In other words, an influence must have at least a 5% effect on a target to be considered
practically significant. Tables 4.5 and 4.6 also show that none of the structural paths
between factors pass the test for weight invariance across all blogs (LM Test Probability
> 0.05). It is readily apparent that the alternative model fits significantly better than the
naïve model (∆CFI = 0.315, ∆χ2 (0) = 415).
FIGURE 4.6
Naïve Factor Model Fitted with Pretest Data
Formal tests of mediation were performed on the model in Figure 3.11 whenever
one construct influenced another through an intermediary construct (e.g., mass
dissemination affects cognitive diversity through cultural tribalism in Figure 3.11). This
109
was done using Hardy and Bryman’s (2004) procedure of bypassing each intermediary,
one at a time, with a direct path and performing a χ2-difference test. In accord with
Cheung and Rensvold’s (2002) recommendation, ∆CFI was also calculated. When
bypassing a mediator construct fails to significantly improve model fit, the validity of
the base model is supported. Baron and Kenny (1986) specified that mediation be tested
by looking at the estimated parameter value of the path bypassing each mediator; values
close to zero show the mediator is needed. The results of mediation tests are summarized
in Table 4.7; all mediation tests support the proposed alternative model.
TABLE 4.7
Pretest Alternative Model Tests of Mediation
CFI and Χ2 Difference Test Test of Significance Relationship
CFI X2 ∆CFI ∆X2
Bypass Parameters
Baseline 0.883 χ2 (78) = 232
Indoctrination
to Flocking 0.882 χ
2 (76) = 232 0.001
2 (2) 0dχ = 0.018 (t = 0.432)
Mass
Dissemination
to Flocking
0.883 χ2 (76) = 230 0.000
2 (2) 2, .5dχ ρ= < 0.014 (t = 0.243)
Idea Seeding
to Flocking 0.883 χ
2 (76) = 230 0.000
2 (2) 2, .5dχ ρ= < 0.104 (t = 1.156)
Indoctrination
to Cognitive
Diversity
0.890 χ2 (76) = 221 0.007
2 (2) 9, .02dχ ρ= < -0.034
(t = -0.972)
Mass
Dissemination
to Cognitive
Diversity
0.883 χ2 (76) = 230 0.000
2 (2) 2, .5dχ ρ= < -0.119
(t = -2.333)
Idea Seeding
to Cognitive
Diversity
0.891 χ2 (76) = 220 0.008
2 (2) 12, .01dχ ρ= < 0.034 (t = 0.437)
110
FIGURE 4.7
Alternative Factor Model Fitted with Pretest Data
111
Even though the fit of the naïve model to the pretest data is fairly good, when
assessed in the context of the small sample size, three of the hypothesized relationships
are both statistically insignificant and too small in effect size for the associated
hypotheses to be considered supported at this stage. There are encouraging indications
though, in the effect sizes and valance polarities of the relationships representing H1, H2,
H4 and H6, where there is good support of the associated hypotheses.
The alternative model is both well fitted to the pretest data and has relationships
between factors that are both statistically and practically significant. Note that the
alternative model suggests explanations for the hypotheses not supported in the naïve
model. The reason why the hypothesized association between idea seeding and cognitive
diversity (H3) was unsupported seems to be because the grassroots influence of need-for-
cognition is required. Similar speculation seems justified for the hypothesized
relationship between mass dissemination and cognitive diversity (H5), the grassroots
influences of cultural tribalism and need-for-cognition seem to be required. Such
speculation also seems justified by the weakness of the hypothesized relationship
between indoctrination and cognitive diversity (H6) even though the hypothesis was
statistically and practically supported. The unsupported hypothesized effect of
reciprocity on cognitive diversity (H7) may be due to an insufficient conceptualization of
reciprocity. It seems the general principle behind the concept is sound, as lagged value
was associated with increased numbers of participating commenters in the alternative
model.
112
In this section of the chapter, two alternative models have been compared with
varied results. It seems reasonable to ponder whether there is another model that would
be superior compared to the models considered. In the next two sections, two novel
methodologies are discussed that purport to find causal models from correlation data.
These methodologies are used to search for better models that are then compared with
the ones already described.
Directed Acyclic Graphs
Directed Acyclic Graphs (DAGs) are similar to structural equation models
(SEMs) in that they are a pictorial diagram of the relationships between variables. The
important difference is that where SEMs are representations of simultaneous linear
equations, DAGs represent the findings of an artificial intelligence algorithm that uses
correlations and metadata about temporal relationships between variables to propose an
underlying causal structure. Swanson and Granger (1997) suggested that DAGs could be
used as a data-driven means of inferring causal structure when little theory is available.
Many alternative algorithms are used to construct DAGs; however the most well-known
are Spirtes, Glymour and Scheines’ (2000) PC search and Chickering’s (2002) Greedy
Equivalent Search (GES). Regular users of DAG algorithms (e.g., Zhang, Bessler and
Leatham 2006) tend to place greater confidence in relationships found by multiple
algorithms over those found by just one.
113
FIGURE 4.8
PC Search DAG
PC Search
The PC search algorithm begins by connecting every variable with an undirected
edge (i.e., a line connecting them in the diagram). Edges are removed where correlation
is either zero or conditional on other relationships between variables. Edges are made
directed (i.e., given an arrow head) in a subsequent step too complex to be described
here. When PC search was applied to the raw correlation matrix underlying that in Table
4.4, the DAG depicted in Figure 4.8 resulted.
114
FIGURE 4.9
Greedy Equivalent Search DAG
Greedy Equivalent Search
The Greedy Equivalent Search (GES) algorithm is a good complement to PC
search because it begins by assuming that no edges connect the variables. It uses a two-
step iterative procedure and a Bayesian analysis of probabilities to converge on a highest
likelihood set of edges between the variables. When GES was used on the raw version of
Table 4.4’s correlation matrix, the DAG in Figure 4.9 emerged.
115
FIGURE 4.10
PC Search and GES Commonality DAG
Figure 4.10 shows the edges found by both the PC Search and GES algorithms.
The edges that match those on the naïve and alternative models are labeled. The findings
of the data-driven algorithms lend support for some of the relationships proposed in the
theory-based alternative model. The inclusion of one of the unsupported hypotheses
from the naïve model (H7) in a mediator role between indoctrination and cognitive
diversity is interesting and worthy of further thought.
Limitations
Barnett and Woelfel (1979) point out that it can be difficult to select the optimal
number of dimensions in the multidimensional space where cognitive objects are plotted.
116
I propose to use four dimensions in this study because as dimensionality increases, the
computational tasks for self-assembling thematic clusters rise dramatically along with
the computer memory required.
The results of this study, while relevant to the broad spectrum of marketing
phenomena, may not be completely replicable in a blog study closer to the core business
focus of mainstream marketing research. As stated in Chapter III, the social forms in this
study are often highly dependent on initial conditions, are path dependent, nonlinear and
only loosely coupled to each other. They often also involve random delays and different
timescales. The more specific in context a future blog study may be, the more likely that
context will be affected by a lesser number of social processes.
The purposive sample of blogs may cause the effects of processes that may be
rare, such as indoctrination, to appear to have greater importance, relative to the other
processes, than is true of the overall blogosphere.
Summary
In this chapter, a means of testing the models proposed in Chapter III was
described and its effectiveness demonstrated with pretest data. In the next chapter all 15
blogs in the dataset are subjected to the testing procedure described here to create a basis
for final determinations about the proposed hypotheses and models.
117
CHAPTER V
RESULTS
Chapter IV described this study’s data set and a multifaceted series of procedures
designed to test the hypotheses and models introduced in Chapter III. The data set is a
diverse selection of 15 blogs, selected using guidelines specified by Kozinets (2002) to
ensure a data set with high information content. The blogs were also selected so as to
cover the full breadth of marketing activities: for-profit and non-profit, organizational
and individual. Blog entries and associated comments were converted into word
networks by Centering Resonance Analysis (CRA). The similarity (or resonance)
between these word networks was measured and used as an input to multidimensional
scaling to locate blog entries and comments as points in a cognitive space. Standard
clustering algorithms were used to find thematic clusters among the points in cognitive
space. The clusters were measured on such dimensions as size and density to derive a
measure of cognitive diversity. Cognitive diversity metrics were then used to segment
blog entry comments into those most influenced by cultural tribalism and need-for-
cognition. Such metrics were also used to segment commenters into idea seeders,
evangelists and flockers. Multi-group structural equation modeling was then used to test
a model and hypotheses that related the joint influences of blog author and commenters
on the diversity of thought expressed in blogs.
This chapter begins by describing the results of conducting the tests and
procedures described in Chapter IV on all 15 of the blogs in the full dataset. On the basis
118
of those results, final determinations are made concerning support for the proposed
hypothesis and models.
Classification of Blog Entry Comments
Chapter III described how blog entry comments can be classified as either
influenced primarily by cultural tribalism or need-for-cognition using the discriminant
model equation 3.1 or, by plotting the positions of comments in the conceptual map
shown in Figure 3.2, derived from equation 3.1. The two comment types were
distinguished, using Chui et al’s (2001) clustering algorithm using all the blog entries
and comments from each of the 15 blogs. Two clusters emerged and cluster membership
was confirmed using McQueen’s (1967) k-means. When a random sample of 500
comments was drawn from each blog in the dataset, pooled and plotted on the
conceptual map of Figure 3.2, the result is as displayed in Figure 5.1 (B). While the
results are not perfectly conforming to the proposed perceptual map regions, the higher
density of data points within the so-called cultural tribalism and need-for-cognition
quadrants suggests support for the proposed topology for blog entry comment
segmentation. It is easy to see that the clustering algorithms merely cut the mass of
comments in the middle. To get a more nuanced insight, it is necessary to look at the
comment frequency histograms below and to the left of the scatter plot (A and C). Even
though the two clusters contain approximately equal numbers of points (50.3% cultural
tribalism), the cultural tribalism cluster is denser, as shown by its smaller footprint in the
scatter plot and taller height in both the individual and collective thought diversity
119
histograms. This observation was confirmed by a statistically significant difference (F =
28.53) between the mean distance between cluster members and their cluster centers.
FIGURE 5.1
Blog Entry Comment Mapping
120
This observation is consistent with cultural tribalism theory; not only are both types of
thought diversity low, but there is less variation even at those low levels. It is also
interesting to note that high and low levels of individual and collective thought diversity
seem to occur together; the two are moderately correlated 0.450 (ρ < 0.001).
FIGURE 5.2
Blog Cultural Tribalism Histogram
121
The histogram in Figure 5.2 addresses the question of whether the blogs differ in
the number of comments classified as predominantly influenced by cultural tribalism.
The white band indicates a 3σ confidence interval around the mean. There is no
statistically significance difference in the level of cultural tribalism between the blogs.
To gain another perspective on the question of whether blogs differ in the
relative effects of cultural tribalism versus need-for-cognition, the underlying
distributions for individual and collective thought diversity were compared in the
paneled histograms of Figure 5.3. In Figure 5.3, the blogs are identified with a number
corresponding to those in the ID column of Table 4.1. It is apparent from Figure 5.3 A
and B that the blogs differ in these underlying distributions, with some seeming to have
skewed (e.g., 10, 11 in 5.2 A), normal (e.g., 1, 6) and bimodal (e.g., 5, 13, 15)
distributions. The skewed distributions all favor the low end of the scale, perhaps
indicating cases where the whole blog is characterized by cultural tribalism. Bimodal
distributions may indicate blogs in transition, where idea seeding is in conflict with
either indoctrination or mass dissemination for ideological dominance.
122
FIGURE 5.3
Paneled Histograms of Thought Diversity
To address these speculations more definitively, paneled histograms of estimated
values for the need-for-cognition and cultural tribalism latent variables were constructed
(Figure 5.4). All the blogs seem to have similarly shaped distributions even though they
are located on different parts of their underlying scale; they seem to have normally
distributed levels of cultural tribalism and positively skewed levels of need-for-
cognition. Since these two constructs share in common the individual and collective
123
thought diversity indicators, and are conceptualized as polar opposites, it may be that the
normal distribution of cultural tribalism is a magnification of the elongated negative tail
of the need-for-cognition distribution. Should this be true, then all the distributions of
Figure 5.3 are bimodal, some merely more apparent than others.
FIGURE 5.4
Paneled Histograms of Cultural Tribalism and Need for Cognition
124
Regardless of whether the distributions of Figure 5.3 are bimodal or not, the evidence
suggests that need-for-cognition is a universally dominant influence in blogs, something
that bodes well for the possibility of high general levels of cognitive diversity.
The parameters of equation 3.1 were estimated from the dataset and the accuracy
of the resulting model tested with a holdout sample. One discriminant function emerged
from the estimation with a characteristic root eigenvalue greater than 1. This function
correctly classified 90.2% of the holdout sample. The coefficients of the discriminant
model are given in Table 5.1.
TABLE 5.1
Comment Discriminant Function Coefficients (Equation 3.1)
Parameter Standardized Unstandardized
β4 (Stdev Individual) 0.457 0.541
β3 (Mean Individual) 0.483 0.565
β2 (Stdev Collective) 0.486 0.553
β1 (Mean Collective) 0.467 0.547
β0 (Constant) -0.042
The only interesting theoretical insight from Table 5.1 is that all the parameters play a
near equal role in separating comments primarily motivated by need-for-cognition from
those motivated by cultural tribalism.
125
Classification of Commenters
Chapter III described how commenters can be characterized as evangelists, idea
seeders and flockers by plotting their positions on a conceptual map (Figure 3.8, derived
from equation 3.2). Idea seeders can be separated from evangelists by high diversity in
thought expression: idea seeders express thoughts about a wide variety of things while
evangelists always express similar thoughts. Idea seeders and evangelists can be
separated from flockers on the basis of differences in boldness: idea seeders and
evangelists are bolder when expressing their thoughts.
When the commenting activity of 7500 commenters, 500 randomly selected from
each of the blogs in the dataset, was pooled, converted to z-scores and demarcated by
Chiu et al’s (2001) clustering algorithm, three clusters emerged (see Figure 5.5 B).
These cluster assignments were confirmed using McQueen’s (1967) k-means algorithm.
Compare Figure 5.5 (A) with Figure 3.7, where idea seeders and evangelists are seen as
occupying opposite tails of a normal distribution along an individual thought diversity
axis. In Figure 5.5 (A) the distribution fits the rough pattern of a normal distribution but
the center is distorted into a sharp peak. It seems as though commenters near the mean
exhibit (or perhaps are drawn into) greater similarity in cognitive consistency (the
cognitive distance between an individual’s thought expressions) than commenters away
from the mean on either tail. This higher density section appears to be within the cluster
labeled “Flocking” or “Flockers” in Figure 5.5 (B). It was predicted that flockers would
be below average in boldness but unrestricted in their overall thought diversity. While
that appears to be generally true, it seems that the “average” flocker perceives and
126
maintains (perhaps through conscious effort or innate tendency to conform) an ideal
level of consistency in their expressed thoughts. Figure 5.6 (B) shows this phenomenon
is consistent across blogs.
FIGURE 5.5
All Blog Commenter Clusters
127
FIGURE 5.6
Paneled Histograms of Boldness and Individual Thought Diversity
This pattern of things in the same vicinity being drawn to a common point is
similar to the physics concept of a fixed point attractor, a stationary point that exerts a
gravity-like force on nearby objects, drawing them closer. Here, evidence of a social
attractor may be present. A full discussion of these results here is out of context;
however, they are discussed in Chapter VI. Figure 5.7 shows that unrestricted thought
freedom, as personified by idea seeders, is consistent across all blogs (the white band is a
2σ (95%) confidence interval around the mean).
128
FIGURE 5.7
All Blog Idea Seeding Histogram
Since the scatter plot of Figure 5.5 (B) holds data from all 15 blogs, it might be
considered indicative of the diversity of boldness and individual thought that
characterizes the whole blogosphere. If that is so, then comparing it with similar plots
using data from one blog might be a qualitative method of assessing cognitive diversity.
Figure 5.8 contains such plots of commenters from all the individual blogs. While most
plots contain the same three basic clusters the size and location of the clusters vary
129
noticeably. For example, it could be strongly argued that blog 6 has the profile of a blog
that reflects the diversity present in the whole blogosphere while such blogs as 12, 13
and 14 are more limited. Note that no comparisons can be made with diversity of
thought in the whole market; that must be deferred to future research.
FIGURE 5.8
Individual Blog Commenter Clusters
130
In Figure 5.8, the clustering algorithms group commenters into similarly located
clusters in all the blogs, with positions roughly as predicted in Figure 3.8. While
interesting, the scatter plots of Figure 5.8 are more instructive when viewed in
conjunction with Figures 5.6 and 5.9. The boldness histograms in Figure 5.6 (A)
generally show bimodal distributions, two separate normally distributed populations on
either side of the mean. The individual thought diversity histograms (Figure 5.6 B) are
consistent in showing one wide normally distributed population, punctuated with a
center peak. The alternative model depicted in Figure 3.11 proposed indoctrination, mass
dissemination and idea seeding to be the exogenous influences driving the other
constructs, Figure 5.9 shows that levels of those influences, as well as flocking, are
roughly normally distributed. Indoctrination seems invariant across blogs, while flocking
changes little. Variance in these influences does not seem to be an obvious explanation
for the variation in commenter cluster configurations shown in Figure 5.8. However,
levels of mass dissemination and idea seeding vary widely and seem logically correlated
with changes in Figure 5.8. Note, for example, that low levels of mass dissemination in
Figure 5.9 (A) blogs 10 and 12 match small mass dissemination clusters in Figure 5.8. A
similar pattern exists for idea seeding in blogs 3 and 9.
131
FIGURE 5.9
Paneled Histograms of Blog Author and Commenter Types
132
In preparation for estimating the parameters of equation 3.2, the commenters
were forcibly classified into the segments of Figure 3.8. Using these topology-based
segment membership assignments, the parameters of equation 3.2 were estimated and
the accuracy of the resulting model tested with a holdout sample. Two discriminant
functions emerged from estimation with data from each of the blogs, correctly
classifying an average of 92.1% of their respective holdout samples. Means of the
coefficients of the respective discriminant models are given in Table 5.2.
TABLE 5.2
Commenter Discriminant Function Coefficients (Equation 3.2)
Function 1 Function 2 Parameter
Standardized Unstandardized Standardized Unstandardized Β5
(Stdev Comment
to Cluster)
0.419 0.671 -0.010 -0.016
β4
(# Clusters) 0.503 0.810 -0.012 -0.019
β3
(Stdev
Individual)
0.035 0.043 0.714 0.897
β2
(Mean
Individual)
0.038 0.046 0.639 0.777
β1
(# Comments) 0.391 0.597 -0.051 -0.078
β0
(Constant) -0.063 -0.033
133
Construct Validity
As stated in Chapter IV, construct validity focuses on the extent to which data
exhibits (1) convergent validity, the extent to which different assessment methods
concur in their measurement of the same construct (also called factorial or measurement
invariance), and (2) discriminant validity, the extent to which the measures of different
constructs are distinct (Campbell and Fiske 1959). The demonstration of factorial
invariance began by establishing baseline confirmatory factor analysis (CFA) results for
each blog dataset. The results were then compared to see if the same indicators loaded
on the same factors. After the factorial structure was found to be the same across blogs,
hypotheses that each free factor loading (each factor’s “first” indicator is constrained to
unity and is thus not free) was equal across blogs was tested by forcing the factor
loadings to be equal and then observing overall fit statistics as well as the LaGrange
Multiplier (LM) Test univariate probability for each factor loading. This probability is a
measure of significance for the LM Test, a test of the hypothesis that a factor loading
varies across groups (in this case, blogs). LM Test probabilities less than 0.05 are
considered statistically insignificant. That is, when the LM Test probability is less than
0.05 the parameter does not have the same value across the blogs. This procedure does
not require equal sample sizes, nor does the test of causal structure invariance described
below.
Invariance of causal structure was tested in a manner similar to the above. The
same structural model was simultaneously fitted to the 15 blog datasets while
constraining all factor loadings, structural path loadings and one error covariance equal.
134
Then multigroup model fit statistics were observed as well as the LM Test univariate
probability for each constrained parameter.
TABLE 5.3
Normalized Estimates of Mardia’s (1970) Multivariate Coefficient
Blog Mardia’s Coeff. Blog Mardia’s Coeff.
AutoBlog 1.562 Gizmodo 3.146
Blog for America 0.857 Google Blogoscoped 2.244
Blog Maverick 2.561 Joystiq 6.435
The Consumerist 3.715 PaulStamatiou 2.359
EnGadget 2.762 Townhall 2.103
The Evangelical
Outpost 4.243
TV Squad 3.204
Fastlane
1.208
The Unofficial Apple
Weblog 3.401
Freakonomics 2.493
Underlying the use of structural equation modeling to perform CFA is a strong
assumption of multivariate normality. If evidence suggests that this assumption is
violated it has important implications for the interpretability of the findings. None of the
indicators in the data set demonstrated significant nonzero univariate skewness (one tail
of the distribution is longer than the other) or kurtosis (variance is due to infrequent
extreme deviations from the mean). Normalized estimates of Mardia’s (1970)
multivariate coefficient, an aggregate indicator of kurtosis, are given in Table 5.3.
Bentler (2005) suggests that only values greater than 5.0 indicate non-normally
distributed data. Another test of multivariate normality is discussed later.
135
The individual blog baseline CFA results showed the same indicators loading on
the same factors as given in Table 5.4. Additionally, as Table 5.4 also shows, all but one
of the 32 core factor construct indicators met Hair et al’s (2005) criterion for statistical
significance (factor loading > 0.4 with a sample size > 200) and 27 met the criterion for
practical significance (loading > 0.5). It must be noted that Hair et al’s criteria, as well as
most of the reference values quoted in this study, applies to evaluating instruments used
to gather experimental data. The circumstances of this study are much different as Hair
et al’s criteria are being used to evaluate a means of gaining insights from non-
experimental, albeit primary, empirical data. Therefore, it is argued that Hair et al’s
criteria must be referenced with caution as the level of control over extraneous variation
that characterizes experiments is absent here. That being understood, the general
conformance of this study’s factor loadings and other metrics to those recommended by
Hair et al. (2005) and Byrne (2006) is worthy of note and indicative of a reasonable level
of performance. The multigroup CFA with factor loadings constrained equal across
blogs had good fit (CFI = 0.884, RMSEA = 0.025, χ2 (3244) = 22,788) indicating that all
blogs have similar, if not the same, factor structures. Table 5.4 also shows that all of the
32 factor loadings exhibited a less than 5% LM Test univariate probability, indicating
that their factor loadings are not the same across the blogs even though the structure
appears to be the same.
Discriminant validity was assessed in two stages: (1) correlations between
constructs, corrected for attenuation, were confirmed significantly less than unity
(Tables C-1 to C-15 in Appendix C contain the correlation matrices for all 15 blogs), and
136
(2) a χ2 difference test was performed between the baseline multigroup CFA model (CFI
= 0.875, χ2 (436) = 35,515) and an alternative multigroup model (CFI = 0.403, χ
2 (464) =
167,750) where the correlation between each latent construct was constrained at unity.
According to Cheung and Rensvold (2002), a large CFI (i.e., greater than 0.01) or χ2
difference demonstrates discriminant validity. It is readily apparent that a ∆CFI of 0.472
and a ∆χ2 of 132,235 with degrees of freedom equal to 28 argue strongly in favor of
discriminant validity.
TABLE 5.4
Confirmatory Factor Analysis, Reliability, Variance Explained and Factorial Invariance
Main Factor and Reliability
Indicator Loading R2 LM Test Probability
Mean Cluster Density -0.651 0.424 0.000
Mean Cluster Radius 0.593 0.352 0.000
Number of Clusters 0.577 0.333 0.000
Cognitive
Diversity
r = 0.711
Percentage of Outliers 0.645 0.417 N/A
Mean Collective Thought
Separation -0.525 0.276 0.000
Stdev Collective Thought
Separation -0.518 0.268 0.000
Mean Indiv. Thought
Separation -0.500 0.250 0.000
Stdev Indiv. Thought
Separation -0.534 0.285 N/A
Mean Commenter Longevity 0.533 0.284 0.000
Cultural Tribalism
r = 0.683
% Repeat Commenters (T - 1) 0.454 0.206 0.005
Mean Clusters Started -0.525 0.275 0.000
Stdev Indiv. Thought
Separation From Cluster
Centroid -0.585 0.342 N/A
Flocking
r = 0.558
Mean Number of Comments -0.518 0.269 0.000
137
TABLE 5.4 (CONT)
Main Factor and Reliability
Indicator Loading R2 LM Test Probability
Mean Clusters Started 0.758 0.575 0.002
Stdev Indiv. Thought
Separation From Cluster
Centroid 0.829 0.687 0.000
Mean Number of Comments 0.748 0.559 0.000
Mean Indiv. Thought
Separation 0.537 0.289 0.000
Idea Seeding
r = 0.852
Stdev Indiv. Thought
Separation 0.598 0.358 N/A
Mean Time Between Entries -0.598 0.357 N/A
Mean Entry-to-Entry Separation -0.578 0.334 0.000
Indoctrination /
Free Thought
r = 0.603 Stdev Entry-to-Entry Separation -0.563 0.317 0.000
Mean Clusters Started 0.468 0.219 0.000
Stdev Indiv. Thought
Separation From Cluster
Centroid 0.608 0.369 0.000
Mean Number of Comments 0.534 0.286 0.000
Mean Indiv. Thought
Separation -0.639 0.408 0.000
Mass
Dissemination
r = 0.690
Stdev Indiv. Thought
Separation -0.464 0.215 N/A
Mean Collective Thought
Separation 0.540 0.291 0.000
Stdev Collective Thought
Separation 0.384 0.148 0.000
Mean Indiv. Thought
Separation 0.405 0.164 0.000
Stdev Indiv. Thought
Separation 0.599 0.359 N/A
Need for
Cognition
r = 0.635
% First Time Commenters 0.539 0.291 0.000
Reciprocity
r = 1.0
Total Number of Commenters
1.0
1.0
N/A
Note: ρ < .05. Reliabilities are calculated from the indicators to each individual factor in
Table 5.4. Pearson “r” is reported instead of Cronbach’s α. N/A in the right hand column
denotes the arbitrary first factor.
138
As has already been intimated, prior to conducting a confirmatory factor analysis,
and the model estimation, it is important to test for multivariate normality in the factor
indicators. This is particularly important in this study as many measures of collective
phenomena have an underlying power-law distribution, as discussed in Chapter II, rather
than a normal distribution. Hair et al. (2005) point out the difficulty in testing
multivariate normality (the joint normality of more than one variable) and suggest that
testing univariate normality is an acceptable substitute, especially if sample sizes are
large. Hair et al. recommend graphical analysis using the normal probability plot as well
as statistical techniques such as skewness and kurtosis (already tested with Mardia’s
(1970) coefficient as shown in Table 5.3). Figure C-1 in Appendix C shows a normal
probability plot for estimated values of each blog’s cognitive diversity latent variable.
These plots show no normality problems in agreement with Table 5.3.
Hypothesis Testing
Blog entry comments were characterized as influenced primarily by cultural
tribalism or need-for-cognition and commenters in each blog were segmented into
evangelists, idea seeders and flockers. Then, the models depicted in Figures 3.10 and
3.11 were estimated and hypotheses tested.
139
FIGURE 5.10
Fitted Core Naïve Model
Prior to fitting the 32 indicator measurements to the models in Figures 3.10 and
3.11, with the blog data, models were translated into simultaneous equations. Then
maximum likelihood estimation (using EQS 6.1) was used, with factor loadings and
paths between factors constrained equal across blogs, to solve the multigroup models
and obtain the results reported in Tables 5.5 (Naïve: CFI = 0.516, RMSEA = 0.126, χ2
(639) = 102,255), 5.6 (Alternative: CFI = 0.810, RMSEA = 0.078, χ2 (649) = 40,464)
and Figures 5.10 and 5.11 (reflective parameters for both models are given in Appendix
D). Tables 5.5 and 5.6 also show that none of the structural paths between factors pass
the test for weight invariance across all blogs (LM Test Probability > 0.05). It is readily
140
apparent that the alternative model fits significantly better than the naïve model (∆CFI =
0.294, ∆χ2 (10) = 61,791).
TABLE 5.5
Multigroup Fitted Naïve Model Parameter Values
Parameter Value Hypothesis Support
Standardized Unstandardized Statistical Practical LM Test
Probability H1 -0.1178 -0.1455 (t = -18.88) Yes Yes 0.000
H2 0.4594 0.6259 (t = 52.00) Yes Yes 0.000
H3 0.1228 0.1257 (t = 22.14) Yes Yes 0.000
H4 -0.0816 -0.0960 (t = -11.98) Yes Yes 0.000
H5 -0.2170 -0.2266 (t = -33.74) Yes Yes 0.000
H6 -0.0165* -0.0171 (t = -2.55)* Yes No 0.000
H7 0.3083 0.4727 (t = 26.92) Yes Yes 0.000
*Not practically significant
TABLE 5.6
Multigroup Fitted Alternative Model Parameter Values
Parameter Value Relationship Support Path
Standardized Unstandardized Statistical Practical LM Test
Probability R1 0.1662 0.1394 (t = 23.33) Yes Yes 0.000
R2 -0.4743 -0.3936 (t = -54.75) Yes Yes 0.000
R3 -0.0017* -0.0012 (t = -0.26)* No No 0.000
R4 -0.2796 -0.1936 (t = -38.30) Yes Yes 0.000
R5 0.3124 0.3269 (t = 36.09) Yes Yes 0.000
R6 0.2357 0.2956 (t = 28.63) Yes Yes 0.000
R7 0.5478 0.3776 (t = 59.25) Yes Yes 0.000
R8 -0.2150 -0.2731 (t = -32.22) Yes Yes 0.000
R9 0.5262 0.8012 (t = 55.39) Yes Yes 0.000
*Neither practically nor statistically significant.
141
FIGURE 5.11
Fitted Alternative Model
142
Formal tests of mediation on the model in Figure 3.11 were conducted whenever
one construct influenced another through an intermediary construct (e.g., mass
dissemination affects cognitive diversity through cultural tribalism in Figure 3.11). This
was done using Hardy and Bryman’s (2004) procedure of bypassing each intermediary,
one at a time, with a direct path and performing a χ2-difference test. Cheung and
Rensvold’s (2002) ∆CFI test and Baron and Kenny’s (1986) test of parameter values
were also conducted. The results are summarized in Table 5.7; all mediation tests
support the proposed alternative model.
TABLE 5.7
Alternative Model: Tests of Mediation
CFI and Χ2 Difference Test Test of Significance Relationship
CFI X2 ∆CFI ∆X2
Bypass Parameter
Baseline 0.810 χ
2 (649) =
40,464
Indoctrination to
Flocking 0.810
χ2 (634) =
40,434 0.000
2 (15) 30,
.02
dχ
ρ
=
<
0.002
(t = 0.136)
Mass Dissemination
to Flocking 0.811
χ2 (634) =
40,423 0.001
2 (15) 41,
.001
dχ
ρ
=
<
0.263
(t = 0.8422)
Idea Seeding to
Flocking 0.810
χ2 (634) =
40,461 0.000
2 (15) 3,
.99
dχ
ρ
=
>
0.143
(t = 1.376)
Indoctrination to
Cognitive Diversity 0.815
χ2 (634) =
39,414 0.005
2 (15) 1050,
.001
dχ
ρ
=
< -0.079
(t = -4.833)
Mass Dissemination
to Cognitive Diversity 0.817
χ2 (634) =
39,118 0.007
2 (15) 1346,
.001
dχ
ρ
=
< -1.811
(t = -4.198)
Idea Seeding to
Cognitive Diversity 0.812
χ2 (634) =
40,101 0.002
2 (15) 363,
.001
dχ
ρ
=
<
-0.076
(t = -0.607)
143
As summarized in Table 5.5, H1 to H5 and H7 are supported as indicated by both
statistical significance and effect size. Statistical significance is not noteworthy in this
study, as the large sample size is assured to raise statistical power enough to make all the
estimated values statistically significant. The sizes of the estimated parameter values
show they meet the level of practical significance arbitrarily set in Chapter IV in every
case except H6.
TABLE 5.8
Summary of Hypotheses and Results
Discussion H1 Cultural tribalism is negatively related to cognitive diversity. Figure 5.10
and Table 5.5 show that H1 is supported with both statistical and practical
significance.
H2 Need-for-Cognition is positively related to cognitive diversity. Figure 5.10
and Table 5.5 show that H2 is supported with both statistical and practical
significance.
H3 Idea seeding is positively related to cognitive diversity. Figure 5.10 and
Table 5.5 show that H3 is supported with both statistical and practical
significance.
H4 Flocking behavior is negatively related to cognitive diversity. Figure 5.10
and Table 5.5 show that H4 is supported with both statistical and practical
significance.
H5 Mass dissemination is negatively related to cognitive diversity. Figure 5.10
and Table 5.5 show that H5 is supported with both statistical and practical
significance.
H6 Indoctrination is negatively related to cognitive diversity. Figure 5.10 and
Table 5.5 show that although H6 was estimated with the right valance (i.e.,
negative sign) and is supported by statistical significance, it fails the test of
practical significance (i.e., parameter effect size > 0.05). However, Figure
5.11 and Table 5.6 indicate that when mediated by cultural tribalism,
indoctrination has the hypothesized negative relationship with cultural
tribalism.
H7 Reciprocity is positively related to cognitive diversity. Figure 5.10 and Table
5.5 show that H7 is supported with both statistical and practical significance.
144
Summary
In this chapter, the procedures detailed in Chapter IV were conducted on the
entire data set. All but one hypothesis (H6) in the naïve was supported; however, the
essence of that hypothesis was supported in the alternative model. The alternative model
of Figure 3.11 was shown to fit the data better than the naïve model of Figure 3.10. In
the next chapter the implications of these findings are discussed. A summary of the
hypotheses tested and their results is given in Table 5.8.
145
CHAPTER VI
DISCUSSION AND CONCLUSIONS
Chapter V described the qualitative and quantitative results of this study. In this
chapter, the research questions introduced in Chapter I are reviewed in the context of the
findings of this study. Then the research implications of this study are discussed, along
with suggestions for future research directions. Finally, the managerial implications of
this study’s findings are discussed and actionable recommendations are made.
Review of Research Questions
Mechanisms Explaining the Expression of Thought
In Chapter III, two models were introduced: (1) a naïve model that sought to
explain the extent to which six social processes influenced cognitive diversity, the
expression of diverse thought, and (2) an alternative model, that on the basis of early
theoretical reasoning, sought to explain one way in which the six processes of the naïve
model might interact. Since processes are dynamic, this study took data snapshots at
blog entry intervals of summary measures (mean and standard deviation) of blog author
and individual activity as well as collective activity under the assumption that processes
can be measured indirectly by measuring their effects: regularity in the cognitive
distance between, quantity and timing of thought expressions. In the following sections
observations of the effects of the six social processes will be discussed.
146
Cultural Tribalism
In Chapter III, cultural tribalism was described as a process of individuals
gravitating to blogs where they can express their views without compromise. Since
individuals tend to temper their self-expression to be accepted by those around them,
people who are highly sensitive to the dissonance resulting from compromised self-
expression can never feel truly comfortable unless they are in the company of those
whose views are similar. Thus cultural tribalism results in bringing like-minded people
together. The same mechanism should also cause individuals to select a blog community
where the blog author espouses views similar to their own.
It was expected, as shown in Figure 3.2, that contexts highly influenced by
cultural tribalism would be characterized by relatively low levels of individual and
collective thought diversity as like-minded people keep discussing the same ideas over
and over. It was also expected, as shown in Figure 3.3 where Mean Commenter
Longevity is an indicator of cultural tribalism, that such contexts would nurture long
term associations manifested in a high proportion of serial commenters. It can be seen
from Table 5.4 that these attributes were reliably (r = 0.683) associated together into the
single latent variable construct denoted cultural tribalism.
It was expected that cultural tribalism would be negatively associated with
cognitive diversity in general (H1). This hypothesis was supported in terms of practical
and statistical significance; however, its association with cognitive diversity was found
to be weaker in the naïve model than in the alternative model (compare the relative
magnitudes of H1 and R1) where it was conceived as being a grassroots mediator
147
between the exogenous influences of mass dissemination and indoctrination on cognitive
diversity.
It was interesting to discover that cultural tribalism seemed to have an influence
in all blogs even though blogs do differ in their relative influence (Figure 5.4 A). Scoble
and Israel’s (2006) concern that cultural tribalism is the central dynamic in the
blogosphere seem unfounded as need-for-cognition seems to exert more than double the
effect of cultural tribalism (compare the relative magnitudes of R8 and R9 in Figure
5.11). It does seem to be, however, the prime influence motivating flocking (compare R5
with R6). As discussed later, flocking seems to have a major presence in the blogosphere
and it may well be that cultural tribalism’s support of flocking is its underappreciated
major consequence.
Need-for-Cognition
Need-for-cognition is widely discussed in the psychology literature as a
motivation that underlies behavior, particularly the extent to which people analyze
others’ attempts at persuasion. In this study, the term is used to denote the counter-
influence to cultural tribalism (recall the discussion surrounding Figure 3.2). The need-
for-cognition motivation drives people to express a diverse array of thoughts as they
discuss and respond to the joint stimuli of blog author entries and the comments of other
blog readers.
As an influence that competes with cultural tribalism, need-for-cognition is
conceptualized as being associated with relatively high levels of individual and
collective thought diversity (see Figure 3.2) as well as a higher inflow of new voices into
148
the blog conversation (refer to Figure 3.3 where “% First Time Commenters” is an
indicator of need-for-cognition). It can be seen from Table 5.4 that these indicators were
reliably (r = 0.635) associated together into the single latent variable construct denoted
need-for-cognition.
It was expected that need-for-cognition would be positively associated with
cognitive diversity (H2). This hypothesis was formally supported by both practical and
statistical significance; indeed, its association with cognitive diversity was found to be
strong in the naïve model (see Table 5.5), and even more so in the alternative model
where it was conceived as being a grassroots mediator (see Figure 3.6 and related
discussion) between the exogenous influences of mass dissemination and idea seeding
on cognitive diversity.
Need-for-cognition is an influence present in all blogs; however, the strength of
its presence varies more widely than that of cultural tribalism (compare Figure 5.4 A and
B). Comparing Figure 5.5 B with Figure 5.9 B seems to indicate a strong correlation
between the strong presence of idea seeding and need-for-cognition, a correlation
confirmed by the strong weight on relationship R7 in Figure 5.11.
Although no apriori hypothesis was advanced regarding the relationship, it was
expected that need-for-cognition would be negatively associated with flocking (compare
the negative polarity of H2 with the positive polarity of H4 in Figure 3.10). However, the
study results show a practically and statistically significant positive relationship between
them (R6 in Figure 5.11). A plausible explanation for this observation follows. In Figure
3.8, flocking was distinguished from the other influences on cognitive diversity on the
149
basis of low boldness rather than constrained thought diversity. Since flocking can have
a full spectrum of thought diversity, cultural tribalism can be associated with the lesser
levels of flocking thought diversity (hereafter called “narrow” flocking) while need-for-
cognition can be associated with the greater (“wide” flocking). These associations are
based on the relationships between cultural tribalism, need-for-cognition and individual
thought diversity depicted in Figure 3.2.
Thresholding and Idea Seeding
Idea seeders are conceptualized as one of three general categories of blog
commenter. They were associated with the behavioral dynamic highlighted in
Granovetter’s (1978) Threshold Models of Collective Behavior because they, being high
in boldness to express their thoughts, often act as an impetus that allows more timid
readers to express similar thoughts, starting a cascade of thought expression as
thresholds of inhibition are overcome.
Idea seeding is therefore associated with high levels of boldness and high levels
of individual thought diversity as their boldness renders them uninhibited in expressing
any thought that comes to mind. Unlike the evangelists of mass dissemination, idea
seeders have no memeset they are devoted to propagating (recall Figure 3.7 and
surrounding discussion); they are merely uninhibited in expressing whatever ideas come
to mind. They are conceived as manifesting their boldness in a higher than normal
cognitive distance between their expressed thought and that of others, and in a higher
volume of comments (Figure 3.4). Since they are the first to broach a new topic of
conversation, they start a higher than average number of new thematic clusters. It can be
150
seen from Table 5.4 that these indicators were very reliably (r = 0.852) associated
together into the single latent variable construct denoted idea seeding.
It was expected that idea seeding would be positively associated with cognitive
diversity (H3). This hypothesis was formally supported by both practical and statistical
significance (see Table 5.5); however its association with cognitive diversity was found
to be weaker in the naïve model, than in the alternative model (compare the relative
magnitudes of H3 with R7 in Figure 5.11) where it was conceived of as being an
exogenous influence mediated by grassroots need-for-cognition (Figure 3.6) in its
influence on cognitive diversity. Need-for-cognition among the mass of blog readers
thus amplifies the influence of idea seeding on cognitive diversity as new thoughts
expressed provoke the thinking of others. The necessity of this mediation was confirmed
by a test whose result appears in Table 5.7 (“Idea Seeding to Cognitive Diversity”)
where bypassing need-for-cognition had an insignificant effect on model fit and a weak
bypass parameter.
Idea seeding seems to have a presence in every blog but the extent of that
presence varies widely (Figures 5.8 and 5.9 B). Although the model of Figure 3.11
portrays idea seeding as an exogenous influence, it is really a manifestation of an
extreme side of the grassroots influences that affect blogs. As already stated, idea
seeders have no agenda; they come to the blog as ordinary readers, have their thinking
stirred and then express those thoughts. From Figure 5.11, it is apparent that their
participation as an influence on the mob is solely responsible for any cognitive diversity
in, and value resulting from, blogging.
151
Flocking
Flockers are conceptualized as one of the three general categories of blog
commenter. They are distinguished by below average boldness in expressing their
thoughts as they are motivated by Maslow’s (1943) need-for-affiliation and a fear of
expressing their thoughts in a manner that alienates others in whose group they wish to
belong. This study seems to indicate that flockers comprise the largest segment of
commenters. As a large population, it stands to reason that they would differ widely in
their willingness to express their true thoughts and in how much they fear exclusion
from the group of their choice. It can be seen from Table 5.4 that the indicators
descriptive of boldness were associated together into the single latent variable construct
denoted flocking, albeit with the lowest level of reliability (r = 0.558) among the
constructs. It is surmised that the one-dimensional nature (i.e., a lack of boldness) of the
construct was responsible for the lower reliability.
It was anticipated that flocking would be negatively associated with cognitive
diversity (H4). This hypothesis was formally supported by both practical and statistical
significance (see Table 5.5); however, its association with cognitive diversity was found
to be very weak in the naïve model. It was anticipated that its influence would be
stronger because limiting the expression of diverse thought robs the collective of the full
array of cognitive resources possessed by its members. However, the observed near
neutral result can be explained by two perceptions: (1) flockers that express nearly their
full scope of thought (wide flocking) counterbalance those that express little of that
scope (narrow flocking), and (2) the system under study knows nothing of potential
152
contribution from its members, only their overt participation. Thus, there is no way for
lost knowledge to be accounted for in the measure of cognitive diversity as when it was
not expressed it never existed as far as the blogosphere is concerned.
Flocking’s positive relationship with both cultural tribalism and need-for-
cognition is interesting. As noted earlier, wide flockers, expected to be more diverse in
their expression, are probably influenced primarily by need-for-cognition whereas
narrow flockers are influenced by their attraction to groups based on cultural tribalism.
Cultural tribalism and need-for-cognition thus mediate the effect of the independent
influences of indoctrination, evangelism and idea seeding on flocking (see Table 5.7,
where bypassing cultural tribalism’s and need-for-cognition’s mediation had no
significant effect on model fit). As it was argued that idea seeding is but an extreme
aspect of the effect of grassroots influences on blogging, so flocking is another aspect of
grassroots influence.
Figures 5.8 and 5.9 (C) seem to suggest that flocking has a ubiquitous presence
in the blogosphere that varies little between blogs. It is wrong then to conclude from
flocking’s neutral influence on cognitive diversity that it has no importance. Flocking is
a manifestation of another form of value gained by blog participants: social belonging.
The model of Figure 3.11 thus portrays flocking as a second dependent variable, that
along with cognitive diversity create value for the community. Just as marketing theory
maintains that consumers have diverse preferences that need to be met, so the grassroots
evolution of the blogosphere has facilitated value creation for at least these two groups
153
of consumers, who to varying degrees substitute knowledge for social affiliation in their
conscious and unconscious choices.
It has already been noted that the individual thought diversity distribution of
Figure 5.5 (A) has an abrupt peak that occurs in all the blogs as shown in Figure 5.6 (B).
The only location in Figure 5.5 (B) where this peak could occur would seem to be the
flocking cluster. Figure 5.5 (A) shows a mainly normal distribution where people near
the mean seem to have been drawn to the mean itself, thus eroding the top of the
distribution and creating a dense pillar. A plausible explanation was suggested in
Chapter V that this is the result of a social attractor, what Manzini (1994) describes as
something that “orients the choices of a multiplicity of individuals” (p. 43). In this case it
is suggested that people near the mean exert effort, consciously or unconsciously, to be
“normal” in the extent to which their expressed thought diverges from their perception of
the community mean. For this to have the observed effect, these people must be very
accurate in their assessment of what “normal” divergence is and what the community
norms are. Since all the blogs used in this study are English language, perceptions of
“normal” may be a part of Western culture, learned adaptations that help individuals live
in greater harmony in Western society. This control of expressed thought divergence is
an integral aspect of how flocking was conceptualized in Chapter II, offering evidence
that this study is truly measuring the flocking phenomenon.
Mass Dissemination
Mass dissemination was conceptualized as one of the two mechanisms used to
spread memes or idea viruses. Mass dissemination may be covertly top-down or
154
hierarchically imposed by an entity with an agenda or it may be another grassroots
influence where the charisma of the meme internally motivates people who embrace it to
spread it. Agents of mass dissemination are denoted evangelists.
Mass dissemination is similar to idea seeding in the boldness of evangelists to
express their thoughts regardless of context. Mass dissemination differs from idea
seeding in the narrow diversity of thoughts expressed by evangelists. The mass
dissemination construct is thus reflected by boldness (a history of being a prolific
commenter expressing thoughts at high variance from the norm, initiating more thematic
clusters) and low diversity of expressed thought. It can be seen from Table 5.4 that the
indicators discussed above were reliably (r = 0.690) associated together into the single
latent variable construct denoted mass dissemination.
It was expected that mass dissemination would be negatively associated with
cognitive diversity (H5). This hypothesis was formally supported by both practical and
statistical significance in the naïve model (refer to Table 5.5). In the alternative model,
mass dissemination’s effect on cognitive diversity was modeled as mediated by cultural
tribalism and need-for-cognition. On the cultural tribalism side, it was reasoned that the
limited view point expressed by evangelists would have an effect similar to
indoctrination in creating an intellectual home for people that embrace the meme.
However, model estimation showed that mass dissemination is negatively associated
with cultural tribalism (R2 in Figure 5.11). Two reasons are offered for this unexpected
finding: (1) a competing meme entrenched in a blog community by officially enacted
indoctrination will actively oppose competing memes introduced by evangelists, and (2)
155
evangelists do not have the intellectual authority to set a blog’s thematic tone. If
evangelists are successful in drawing converts, they must migrate to other more
agreeable blogs or start their own blogs thereby becoming indoctrinating blog authors.
On the need-for-cognition side, it was reasoned that blog readers could not
readily identify a new evangelist, so evangelists would be mistaken for idea seeders and
thereby initiate grassroots interest, increasing evidence of need-for-cognition in action.
However, it may be that evangelists are too zealous and bore blog readers, causing them
to withdraw from participation. Fortunately, in relative terms, evangelists seem less
influential than true idea seeders and do not so affect the discussion as to dissuade it
completely.
Indoctrination
Indoctrination was conceptualized as the second mechanism used to spread
memes. Indoctrination is the only overtly top-down or hierarchically imposed influence
proposed to be acting on blogs. It is seen as being the means of influence possessed by
the blog author, exercised in the writing of blog entries that may set the thematic tone for
the blog and establish the norm for thoughts expressed.
Indoctrination as a construct is similar to the individual thought diversity
construct applied to commenters. It is indicated by the mean and standard deviation of
the cognitive distance between blog entries with the assumption that the more similar the
blog entries, the more intense the indoctrination. It is also indicated by the mean time
between entries, under the assumption that greater indoctrination intensity will be
accompanied by more frequent blog entries. As the indoctrination concept implies intent
156
to persuade, the absence of such intent must also be conceptualized. Since
indoctrination, by definition, discourages critical thinking, its absence must embrace
such thought and is therefore denoted “free thought” in the blog author thought diversity
spectrum of Figure 3.5. It can be seen from Table 5.4 that the indicators discussed above
were reliably (r = 0.603) associated together into the single latent variable construct
denoted indoctrination / free thought.
It was expected that indoctrination would be negatively associated with cognitive
diversity (H6). This hypothesis was not supported by either practical or statistical
significance in the naïve model (refer to Table 5.5). In the alternative model, its effect on
cognitive diversity was seen as mediated by cultural tribalism and need-for-cognition,
the indoctrination side of the spectrum was thought to be more associated with cultural
tribalism (R1 in Figure 5.11) while the free thought side was thought to be associated
with need-for-cognition (R3). Model estimation revealed that while indoctrination was
associated with cultural tribalism (R1), it had neither practical nor statistical effect on
need-for-cognition (R3). It is concluded that the cognitive diversity expressed by the blog
author has only a one-sided impact on the cognitive diversity expressed by the blog
community. It seems to be only able to create an intellectual home for people who agree
with the blog author’s views and those who flock around them. The free thought side of
blog entries may however be a signal of tolerance for diverse viewpoints that, while
undetected by this study, are perceived by readers and result in elevated levels of idea
seeding and wide flocking that, through need-for-cognition, expand cognitive diversity
(R9).
157
Reciprocity
Reciprocity was conceptualized as both a psychological and an economic
influence. Blog readers who gain value from the blog conversation are envisioned as
gradually accumulating a sense of obligation to contribute their own insights. Every
individual has a different threshold of tolerance for the cognitive dissonance resulting
from unreciprocated benefit. However, when that threshold is exceeded, individuals will
write comments contributing their own knowledge to the collective, thereby repaying the
community.
In the naïve model, reciprocity was conceived as being influenced by prior value
created by the blog conversation and manifested by an increase in commenters. With
only one indicator, the construct cannot be viewed as having much content validity;
therefore, it was not included in the alternative model. However, prior value was
included as an influence that might be associated with blog authors focusing on popular
themes, and in so doing appear to be indoctrinators, and with an increase in the number
of commenters.
It was posited that reciprocity will be positively associated with cognitive
diversity (H7), a hypothesis that was both statistically and practically supported in the
naïve model (Table 5.5). In the alternative model, the influence of reciprocity’s implied
motivator, prior value received, was shown to be statistically and practically associated
with indoctrination and the number of commenters (see Figure 5.11). However, even
with these supportive results, it must be acknowledged that the hidden internal nature of
reciprocity prevents a complete accounting of its influence on cognitive diversity, and
158
indeed its role in shaping the behavior observed in the blogosphere as a whole, to be
gained though ethnography. As a result, it cannot be said that the full association
between reciprocity and cognitive diversity was properly assessed here.
Processes that Expand Cognitive Diversity
The preceding discussion can be summarized in concluding that the key to
achieving cognitive diversity is inspiring need-for-cognition in the blog community, and
such need-for-cognition is primarily associated with the presence of idea seeding, that is,
free thinking and uninhibitedly expressive commenters. Idea seeding and need-for-
cognition is a grassroots stimulus-response pair that may indeed have a reciprocal
relationship over time: idea seeding invokes need-for-cognition that generates thoughts
that when expressed to the community results in the seeding of more ideas. Such a
dynamic can be a proverbial perpetual motion machine of thought and cognitive
diversity generation.
Processes that Limit Cognitive Diversity
It seems that cognitive diversity may be adversely affected by indoctrination,
mass dissemination and cultural tribalism. As discussed above, there is evidence to
suggest that mass dissemination reduces cognitive diversity by dampening need-for-
cognition. The thematic repetition of evangelists may exhaust the interest of the
community, causing readers to become uninspired and docile. Blog authors, whether
motivated by an indoctrination agenda or by the desire to feed the community more of
what it appears to enjoy, who limit the thematic content of their posts encourage the
congregating of those who have a narrow interest in that theme, inspiring them to repeat
159
back the same ideas. Similar to the feedback mechanism posited to exist between idea
seeding and need-for-cognition, a blog author who has nurtured a community
characterized by cultural tribalism may, if monitoring the community, exacerbate the
situation by adopting a populist stance in selecting themes for blog entries. It must be
questioned though whether such communities have long term viability; since boredom is
a universal human propensity, one would expect that eventually interest in communities
dominated by cultural tribalism would wane unless they could attract a steady inflow of
new adherents. The gradual decline and disappearance of cultural tribalism communities
may also be a source of reduced cognitive diversity.
Most Important Processes
The resilience of the blogosphere seems strong since influences that seem to
expand cognitive diversity are greater than those that seem to reduce it. It is interesting
to note that indoctrination, the only top-down influence is exclusively associated with
reducing cognitive diversity while grassroots emergent processes are behind the
influences that expand it. Even what may be well-intentioned top-down attempts to give
the community what it wants (prior value feedback to indoctrination), seem only to serve
to damper the expression of diverse thought.
It must be remembered though, that this study narrowly focused on investigating
influences that affect the expression of diverse thought under the assumption that the
value that accrues from such cognitive diversity is of most interest to companies because
they want a balanced and true perception of their image. Other types of value, seemingly
of greater interest to other stakeholders, have been briefly mentioned: value from
160
satisfying need-for-esteem by exercising self-expression (through cultural tribalism),
from satisfying need-for-affiliation (through flocking), and from satisfying need-for-
cognition. It is apparent that the pursuit of these other forms of value differently affects
cognitive diversity. Striving to satisfy need-for-cognition seems to positively influence
cognitive diversity while the other pursuits seem to restrict it. Maslow (1943) leads to
the conclusion that need-for-affiliation is a more basal and compelling need than the
others. As a result, it must be acknowledged that by focusing its inquiry on processes
that affect the filling of cognitive needs, this study may not have examined the major
influences in the blogosphere.
Implications for Research in Marketing
In Chapter I, it was stated that this study sought to make substantive and
methodological contributions to the field of marketing. In this section, the nature of these
contributions is described.
Substantive Contributions
It must be acknowledged that this study was largely exploratory in nature. Its
main contribution is an early-stage detection of patterns and phenomena that require
further investigation in order for greater confidence to be placed in the findings. That
caveat stated, this study does introduce some unique concepts to the marketing literature.
One of those contributions is empirical evidence supporting the concept of flocking as a
factor in group cognition. While this idea was proposed by Rosen (2002), it has not been
hitherto empirically demonstrated in a socio-cognitive context. Additionally, this study
introduces the concept of cultural tribalism into the marketing literature as distinct from
161
groupthink and group polarization. Cultural tribalism is a grassroots, long term
phenomenon resulting from a desire for uncompromised self-expression, whereas,
groupthink occurs because of an externally imposed requirement for consensus and
group polarization occurs from people banding together to exercise greater influence.
This study adds incrementally to the literature investigating what motivates
individuals to express their thoughts, as a form of word-of-mouth, in a public forum. It
proposes and provides evidence for a reason why word-of-mouth might not reflect the
true thoughts of participants: need-for-affiliation may cause some blog participants to
compromise the expression of their thoughts and opinions to maintain social acceptance.
It also introduces need-for-cognition as a word-of-mouth motivator. Furthermore, it
evidences the presence of three distinct types of word-of-mouth participant (flockers,
evangelists and idea seeders) and a model that shows how these types interact in the
same context.
This study also contributes to the small body of literature that explores marketing
phenomena from a complexity science perspective; that is, as primarily influenced by the
grassroots actions of consumers rather than the top-down efforts of companies to drive
consumers into desired behaviors. A common perception of companies is that a blog is
merely another tool of persuasion and image construction. This study presents evidence
to suggest that the readers, through the comments they write, have much greater power
to provoke the thinking of other consumers than the blog author. The social process
theory focus of this study also makes an unusual complexity science contribution to the
literature as it investigates how known processes interact to create the complex behavior
162
of a dynamic context, whereas most marketing studies focus on single processes and
simpler contexts.
This study also adds to the small body of memetic studies in the marketing
literature. The concept of memetics, or ideas that spread like viruses with unusual ease,
has intrigued researchers interested in advertising and persuasion. Unfortunately, it has
been difficult to test models of memetic propagation because they involve unconscious
cognitive processes. While this study does not provide definitive proof of the validity of
the memetic concept or its theories of propagation, its fledgling detection of similarity in
conversational themes and their spread through a group of people known to be in contact
offers support for the proposed mass dissemination and indoctrination concepts.
One of the most interesting contributions of this study is the evidence provided
for the finding of a social attractor in the distribution of individual thought diversity of
Figure 5.5 (A). Like memetics, the concept of social attractors is difficult to validate
because it assumes the operation of social processes that are difficult to empirically
detect. This study documented a strange pattern in what was expected to be a normal
distribution. The social attractor model is a plausible explanation for the pattern
observed, that while inconclusive, certainly provides grounds for further investigation.
Watts and Dodds (2007) challenged the “two-step flow” model of
communication advanced by Katz and Lazarsfeld (1955) and widely embraced by both
the marketing (e.g., Van den Bulte and Joshi 2007; Vernette 2004) and communications
(e.g., Weimann 1994) disciplines. The “two-step flow” model describes information
flowing from mass media sources, through a small number of influential “opinion
163
leaders,” to the public at large. Watts and Dodds (2007) demonstrated through computer
simulations that such information flows depend, not on a few influential individuals (the
classic “influentials hypothesis”), but on developing a critical mass of “easily influenced
individuals” (p. 441) to begin a cascade of information flow with “easily influenced
people influencing other easily influenced people” (p. 447). Watts’ and Dodds’
computational model is persuasive but offered without empirical support.
This study provides some empirical support for the Watts and Dodds (2007)
model, in that it demonstrates the superiority of grassroots mechanisms in propagating
ideas. Both studies draw on Granovetter’s (1978) concept of cascades in collective
behavior overwhelming individual behavioral thresholds to support their theories. It
could be argued that flockers, the seeming majority, are a class of easily influenced
people, as their spreading of ideas is less the result of reasoned conviction than the need
for social acceptance. However, this study also demonstrates the influence of idea
seeders, individuals whose novel ideas frequently start cascades of thinking. Idea seeders
should be considered a hitherto unrecognized class of influential. They act, not as regular
intermediaries between established idea creators and the masses, but as decentralized
and intermittent introducers of diverse ideas who influence the thoughts of others. Some
idea seeders are observed to be more reliable in their influence than others. These idea
seeders may be the classically conceptualized opinion leaders identified by prior
research (e.g., Katz and Lazarsfeld 1955). This phenomenon seems to warrant an
expansion of our conceptualization of what the influentials hypothesis refers to.
164
This study’s empirical findings seem to support the conceptualization of a
complex environment that allows the simultaneous operation of the influentials and
critical mass hypotheses. With regard to the critical mass hypothesis, this study’s
cultural tribalists may be a first wave of early converts to an ideology, while flockers act
as a subsequent, easily influenced early majority that create the critical mass necessary
for mass dissemination. Alternatively, perhaps diffusion begins with the attention of the
need-for-cognition curious, who give a new idea enough visibility for it to attract the
attention of easily influenced flockers who then create a critical mass. It seems more
reasonable, given the history of research that supports the influentials hypothesis, to
consider that both dynamics coexist in a complex environment, rather than to consider
one model supplanting the other.
Methodological Contributions
As detailed in Chapter II, there is a long legacy of research in marketing that has
investigated and attempted to model cognition. This study adds to the marketing
literature a novel integrative application of textual analysis, network theory,
multidimensional scaling and cluster analysis that creates cognitive maps with greater
theoretical transparency than competing methodologies employing neural networks.
The methodology used by this study has garnered interesting, albeit embryonic,
insights that argue its use in other circumstances where text is used to infer cognition.
The methodology is a good quantitative complement to the qualitative analysis of text as
each cognitive object in multidimensional space can be linked back to its underlying
word content, matching cluster patterns to identifiable themes and even theme transitions
165
to changes in a group’s cognitive structure (how knowledge objects were connected
together).
One contemporary theme in the communications literature is discussing the
measure of persuasion effectiveness. Particularly relevant to this study is measuring
attitude toward a message. Often the literature discusses whether measuring perceived
effectiveness (PE) before launching a costly campaign, is indicative of the actual
effectiveness the campaign would have (e.g., Dillard, Weber and Vail 2007). Perceived
effectiveness is generally assessed by surveying a group of message recipients and
asking them whether they find the message persuasive (Dillard et al. 2007, p. 617).
Seeding a candidate message in the form of a reader comment to one or more blogs
where company products are discussed, and using a methodology similar to that
employed in this study to monitor how its content is embraced by subsequent
commenters, may prove to be a more reliable means of testing effectiveness than
surveying PE.
Peterson (2005) posited “with rare exceptions, answers to questions asked in
consumer behavior research studies … tend to be constructed rather than … directly
retrieved from memory” (p. 352). He called for researchers to augment self-reports with
alternative sources of data, including behavioral observation. Monitoring customer
generated media (e.g., blogs and forums) may gain equal footing with (or even displace)
the collection of survey-based respondent driven data as an observational, and thus more
objective, data collection method. This study may be a step in the direction of giving the
research community confidence to do that.
166
This study also contributes to the literature about research methods by being one
of the few large-scale studies using multiple populations. It is thus an example of an
approach that could be taken by those who seek to do a similar study.
Directions for Future Research
Recognizing synonyms in Centering Resonance Analysis’ calculation of
influence (repeated words that tend to connect ideas within a body of text) and resonance
(comparing bodies of text on the basis of the same words used with similar influence)
would be a powerful enhancement to the basic algorithm. However, that enhancement
brings attendant complications as many words in the English language have different
meanings even though they are spelled the same. Work on this enhancement is deferred
to future research.
There seems to be some potential for further insight by examining and analyzing
the data from a time series perspective. To this end the following research questions
merit investigation:
1. Do blog participants grow in the diversity of their thought expression?
2. How do cognitive diversity and the principle constructs (e.g., indoctrination,
mass dissemination and flocking) develop over time, starting with the blog entry
and updated as each comment is added?
3. Is there trending in the major constructs as more blog entries are added to a blog?
4. How do blog communities evolve from inception to present with respect to
membership size, and changes in relative dominance of cultural tribalism and
need-for-cognition as motivating influences?
167
Progress in these areas might also reveal latent variable construct indicators on
additional dimensions that would enhance the reliability and content validity of this
study.
Managerial Implications
Statistically Improbable Words
In Chapter I, companies were portrayed as better able to perform image
management with feedback from the blogosphere. Centering Resonance Analysis (CRA)
has been demonstrated to be a means of identifying the most important keywords
associated with a corporate image. Furthermore, the word network produced by CRA
preserves the context in which the most influential words were used. This capability
could be enhanced by looking for statistically improbable words (SIWs) among the most
influential words demarcated by CRA. These are words that occur much more often than
other words associated with a company, or a company and all its competitors in a
market. The SIW concept is conceptually derived from Amazon.com’s (2008)
Statistically Improbable Phrases (SIPs), the most distinctive phrases inside a particular
book, relative to phrases found in all books. Amazon uses these phrases to hint at unique
themes inside a book that might better separate it from others in the same genre.
Similarly, if this concept is applied to CRA influential words, it may better identify the
most unique words descriptive of corporate image. A company may also be able to see
whether key words from marketing communications are making an impact with the
public.
168
Does Diversity of Thought in the Blogosphere Echo that of the Market?
This study began by asking a very general research question and then proceeded
to answer more specific questions that would hopefully lead to eventually answering the
original question: How do companies know that the diversity of thought they see in the
blogosphere reflects that present in the overall market? Preliminary evidence has been
presented to suggest that a majority of blog commenters (i.e., flockers) compromise the
expression of their true thoughts to gain social acceptance. This study proposes a means
of identifying the most extreme of such people (flockers near and below the mean of
individual thought diversity). Companies who monitor blogs for image insights may be
justified in assigning less weight to image indications gained from extreme flockers as
these commenters may not be expressing their true thoughts, merely thoughts that are
intended to maintain social acceptance.
It remains uncertain whether the population of blog commenters is a
representative sample of the whole market, as this study reveals no profile of cognitive
diversity in the entire market. The only way a conclusion could be made is if it is
assumed that all the measures used in this study (e.g., individual and collective thought
diversity) conform to a certain distribution (e.g., normal) in the overall market. For
example, in Figure 5.5 (A) individual thought diversity is depicted as generally normally
distributed, but with a sharp discontinuity at the mean. It may be assumed that this
attribute is normally distributed in the whole market and thereby conclude that the
population of blog commenters does not represent the whole market (because of the
sharp discontinuity). However, it may be that the whole market of all consumers exhibits
169
a similar discontinuity, supporting the conclusion that the population of blog
commenters accurately represents the whole market. Theoretical argument cannot
resolve this uncertainty as reasons can be offered for both assumptions. It is
recommended that, until an alternative and independent means of corroborating this
study’s results is performed, generalizations remain tentative. Image insights gleaned
from blog monitoring should be examined, but recognized as possessing some degree of
uncertainty yet to be fully understood.
Evidence was presented to suggest that blogs whose content is characterized by a
single theme attract a readership with similarly narrow interests. As Scoble and Israel
(2006) cautioned, such cultural tribalism cannot be a source of balanced image insights.
The nurturing of such communities will be exacerbated if the blog author recycles blog
entry themes because they were popular in the past. If a diverse-thinking blog
community is desired, then it seems that monitoring the proportion of idea seeders is
important. Collectively, the blog populations studied here had an average of 15% idea
seeders. If a blog has substantially less than that, cultural tribalism has probably become
the dominant motivation among blog participants. It seems that idea seeders are nurtured
if blog authors introduce diverse themes into their conversation starting entries.
Blogs as a Tool of Influence
This study’s contribution to the reputation-building literature only addresses the
question of whether blogs might be used as a persuasive tool to alter reputation. This
study identifies a class of blog participant, idea seeders, whose diverse thought
expression seems to reliably spark interaction among a wide variety of blog participants.
170
It is thought that this interaction is primarily of a sensemaking nature, participants trying
to determine where novel thoughts fit into their existing cognitive structures. However,
blog participants who generally express ideas on the same themes, denoted evangelists,
whether or not they have an intent to persuade, seem only to provoke conversation
among individuals with a history of interest in those themes. It seems then, that if idea
seeders begin a practice of repeating ideas on the same theme, then their influence as
conversation starters would diminish. This study observes this behavior pattern in the
community’s reaction to the arrival of new evangelists. The blog community cannot tell
their true nature in the beginning and treats them as idea seeders. However, as their true
nature is gradually revealed, they lose their ability to provoke diverse thought expression
and thereby lose much of their influence in the community.
This study suggests two complementary interventions a company could make in
the blog conversation to ensure it gets a more accurate perception of construed image, its
reputation among consumers:
a) Tactical manipulation. This action is a variation on Dellarocas’ (2006) strategic
manipulation where firms post anonymous positive messages about their
products to opinion forums. However, in this case the intent is not to persuade
readers of the value of a product to induce purchase, but to give silent, less bold
readers the courage to reveal their similar thoughts and thus start a cascade that
reveals a hidden sentiment segment. To execute this tactic a company introduces
a group of covert and seemingly independent evangelists. Each evangelist would
initially be received as an idea seeder and have a limited opportunity to have the
171
community to consider their ideas. This limited period is the evangelist’s primary
opportunity to embolden silent blog community members into revealing their
perspective. If each evangelist expresses thoughts narrowly on a common theme,
the illusion of natural grassroots action may be lost. However, if each evangelist
frames their message from a slightly different perspective, expressing
understated interest or tentative support, they may be able to create the illusion of
sufficient widespread agreement, Cialdini’s (1984) social proof or band-wagon
effect, to draw out enough other community members that the critical mass
needed for a Watts and Dodds (2007) style information cascade is reached. It
should be noted though that this tactic may primarily influence flockers, those
motivated by a need for social belonging, and as a result, it may be superficial in
its indications.
b) Blog author encouragement. This study suggests the blog author is more
influential with cultural tribalists (i.e., the already persuaded) than those
motivated by need-for-cognition. From Figure 5.8 it seems that some blog
authors are able to act on the fine line between maintaining the reputational
beliefs of the tribe and encouraging a free and open dialogue. An atmosphere
welcoming free thought should attract and maintain a diverse-thinking population
of readers. As long as this diversity is present, occasional tactical acts of
manipulation may be effective in revealing new segments of product enthusiasts.
Godes et al. (2005) proposed a framework with four non-mutually exclusive
ways that a company could be involved with online social interactions between
172
consumers: (1) Observer: a company uses their observation of interactions among
consumers as a learning opportunity; (2) Moderator: a company encourages consumers
to interact by establishing a blog or other form of online community; (3) Mediator: a
company takes control of the information contained in online word-of-mouth and
disseminates it as it desires (e.g., removing blog comments it sees as undesirable); and
(4) Participant: a company enters into social interactions with consumers by acting as an
anonymous consumer. Plainly, the latter two behaviors have the potential to manipulate
consumer perceptions. Dellarocas (2006) analyzed the participant scenario from the
perspective of it being a competitive strategy among firms. He concluded that when
firms compete on the basis of manipulating the consumer, it is a costly distraction from
making more substantive market offering improvements. Mayzlin (2006) also analyzed
the participant scenario from the same perspective and found that “firms spend more
resources on promoting inferior products,” (p. 155) because “they don’t get the free
publicity from legitimate chatters.” (p. 161)
Balasubramanian and Mahajan (2001) note that one of the primary advantages of
community is that it allows individuals to build a reputation that streamlines exchange
because trustworthiness does not have to be ascertained prior to every transaction. Bart
et al. (2005) note that community features [e.g., weblogs] are a factor driving trust in
websites. They note that “shared consciousness and a sense of moral responsibility and
affinity enhance the consumer’s level of trust” (p. 136) and may make consumers more
confident in accepting reputational information from online communities. Therefore, any
173
corporate participatory behavior that undermines trust robs blogging of some of its
ability to create positive outcomes for both firms and consumers.
This study’s findings suggest that, since they tire of evangelists, blog participants
have a resistance to sustained corporate manipulation. They may derive this resistance
by matching a commenter’s user name with memory of the content of their past
contributions. Alternatively, they may simply remember the ideas heard without
reference or association to who said them. The latter alternative is similar to the SIR
epidemiology model (Kermack and McKendrick 1927) where individuals are in one of
three states: susceptible, infected, and recovered. Variants of this model have been
applied to ideas and influence in the marketing literature (e.g., Van den Bulte and Joshi
2007). If blog participants become resistant to repeated ideas without reference to
source, sustained corporate manipulation, even in the guise of different evangelists, is
likely to prove futile.
Alternative Sources of Value
Companies should be mindful that people participate in blog communities for
reasons that might undermine thought diversity (e.g., need-for-affiliation). There may be
value in servicing those needs in a corporate blog even though commenters might not be
providing valuable insights into image or directions for product development. Such
communities may enhance the loyalty of customers and have a positive effect on sales.
174
Managerial Implications of Research Findings Linked to Research Questions
In the Review of Research Questions section at the start of this chapter, the
research questions were reviewed in light of the study’s findings. In this section, the
managerial insights of the research are linked to the research questions of the study:
1. What mechanisms explain the expression of thoughts by individuals in a blog?
Figure 5.10, the fitted naïve model, shows that reciprocity and need-for-cognition are
the most important direct influences on cognitive diversity. Figure 5.11, the fitted
alternative model, shows that idea seeding is the primary influence on need-for-
cognition. Therefore, reciprocity and idea seeding seem to be most relevant in
determining whether or not comments are made, and if so, when they are made. This
finding presents a challenge as blog authors may feel pressured to induce reciprocity
by consistently posting content known to be of high value to their community and
thereby fall into indoctrination behavior. Blog authors should be cognizant of what
themes interest readers, but should be more focused on expanding the variety of
relevant themes discussed in the blog, creating an atmosphere of free thought to
encourage idea seeders. Let the value of spontaneous, albeit happenstance,
contributions of a large population of idea seeders and wide flockers be the engine
that induces reciprocity.
2. What processes expand cognitive diversity, the expression of diverse thought? The
influences on cognitive diversity with the largest positive effect sizes were
reciprocity and need-for-cognition (Figure 5.10), while need-for-cognition was
primarily influenced by idea seeding (Figure 5.11). As discussed in response to
175
question 1 above, the most important influences in blogs are also the most positive in
supporting diverse thought. Reciprocity and need-for-cognition seem to work
together to create self-sustaining conversations that attract wide interest and diverse
contributions.
3. What processes limit cognitive diversity? Mass dissemination (evangelists repeating
the same ideas widely to influence those with the lowest thresholds of resistance),
flocking (particularly, narrow flocking where blog participants seek social
acceptance in groups characterized by cultural tribalism) and cultural tribalism
(homogeneous groups formed to satisfy members’ desire for self-expression in a
non-judgmental, albeit echo chamber, context) are the primary influences reducing
cognitive diversity. The implications of the findings relevant to these influences are
discussed next:
a) Evangelists are often received as idea seeders when they first start speaking
their message; however they are quickly disregarded as they repeat the same
ideas. Evangelists may initially expand cognitive diversity before they are
recognized. This presents an opportunity for companies trying to test for the
presence of readers who agree with the evangelized message but are too timid
to express it. It is anticipated that converts to an evangelized message will
leave a blog to found their own, or migrate to a blog that embraces those
ideas.
b) Although narrow flockers, as the most timid and least independent content
contributors, are unlikely to introduce ideas leading to breakthrough products,
176
they may be a very profitable consumer segment if they see their group
acceptance as tied to imitating the product purchases of others. Their
profitability may derive from a magnification of sales effort, where sales
effort expended on influencing more discriminating consumers may trigger
multiple imitative sales. Narrow flockers are also unlikely to be demanding
consumers as long as their need for affiliation is met.
c) Cultural tribalists are the diehard loyalists whose complete satisfaction should
be least susceptible to fickle change. However, companies should not assume
that there is no competition for the loyalty of these community members.
Companies should use the opportunity the passionate participation of the
tribe will provide to ensure their basic needs are not overlooked.
4. What is the relative importance of these processes? The most important influence in
the blogosphere is that idea seeding invokes need-for-cognition dynamic discussed
above. The secondary influences of indoctrination enabling cultural tribalism, free
thought supporting need-for-cognition, and the imitation of both groups influencing
flocking are also very important. Figure 5.4 shows that in all the blogs used in this
study, cultural tribalism and need-for-cognition are able to co-exist to varied degrees
even though they are conceptualized as opposing influences. It is proposed that the
key to this balance is the blog author recognizing the ideas for which the tribe
demands active support (through indoctrination), and then create an atmosphere of
free reign for as wide as possible an array of other ideas that do not contradict those
actively espoused. It is apparent from Figures 5.4 and 5.8 that not every blog is
177
equally successful in achieving a balance that assures high cognitive diversity.
However, companies whose own blogs support little diversity can always monitor
other blogs frequented by their target markets to round their perspective of image
and market satisfaction trends.
Public Policy Implications
Although, public policy issues are not the focus of this study, it is readily
apparent that measuring the diversity of thought in a population around a public policy
issue would be of great interest to legislators and pollsters. All the managerial
implications of monitoring image could apply to any public policy issue, as could this
study’s speculations of how tactical manipulation could be used to reveal points of view
held by non-expressive consumer segments. Three of the blogs used in this study are
host to political discussions and it seems that the same kinds of participants (i.e., idea
seeders, evangelists and flockers) explain the expression of thought in political blogs as
those that are more commercial. A full exploration of this issue is deferred to future
research.
Limitations
While the findings of this study have important implications for marketing
practice and research in marketing, certain limitations must be borne in mind. For
instance, the concepts that the study tries to measure are inherently complicated. This
study takes an admittedly simplistic, albeit good-faith, approach to representing them as
constructs. If the explained variation (R2) estimates were corrected to reflect this
178
uncertain content validity, their values would certainly be much lower. Not only does
content validity need to be assessed and raised, but the reliability of most of the
constructs needs to be increased as well.
Also, the blogs in the dataset are all from the leftmost side of Figure 2.3, the most
popular blogs. The profile of blog populations presented in this study may not reflect
that of less popular blogs, the blogs that collectively comprise the majority of content in
the blogosphere.
This study began by asking: How do companies know that the diversity of
thought they see in the blogosphere reflects that present in the overall market? It might
be argued that the scattering of points in Figures 5.1 (B) and 5.5 (B) are sufficiently
symmetric and well-distributed throughout their respective spaces to show that the full
spectrum of thought has been captured. However, these scatter plots collectively only
cover a three-dimensional space (boldness, individual and collective thought diversity).
If more dimensions were added, Bellman’s (1961) curse of dimensionality would come
into play due to the exponential increase in volume. As volume increases, the distance
between objects located in the space becomes greater and the population density
becomes sparser. Thus, the number of data points required to adequately sample the
space becomes larger. The cognitive space of the overall market is sure to be an
unmanageably high-dimensional space. However, with further research it may be
possible to identify its principal dimensions, define a manageable cognitive space and
map what is sure to be an even larger sample of blog entries and comments to it, and
179
thereby claim to have adequately approximated the extent to which the blogosphere
represents the diversity of thought in the market.
Summary
This chapter discussed the major findings of how the six social processes
investigated in this study were observed to be associated with cognitive diversity, the
expression of diverse thought in blogs. Idea seeding, or the comments of the boldest and
most diverse thinkers, was observed to have the most powerful influence on overall
cognitive diversity because of its ability to provoke the thinking of others, inspiring them
to become commenters too.
The most surprising findings were: (1) the intellectual content of blog authors’
posts is not as influential as the intellectual contributions of the commenters, however
the post’s affective content is thought to be very important in creating a collaborative
atmosphere for commenters; (2) flockers whose thought diversity is near the mean, seem
able to estimate the average level of flocker thought diversity. This estimate appears to
be a social attractor, a cognitive magnet for flockers who aspire to be average in the
diversity of their expressed thought; and (3) idea seeders may be a new class of
influential consumers, distinct from classical opinion leaders in that, while they exert
power as a class, they may hold little influence as individuals. However, the classic
opinion leaders may be those idea seeders who have more consistent influence.
This chapter also discussed the contributions of this study to the marketing
literature. This study explores a new subject and introduces a new methodology to the
field of marketing. However, being largely exploratory in nature, it tended to find
180
evidence for fruitful paths of further inquiry rather than definitive conclusions. Finally,
this chapter made recommendations to marketing managers who are thinking about
using blogs to gain image insights and as a tool for influencing image perceptions. It was
recommended that caution be exercised in using insights gained from blogs because
participants often compromise the expression of their thoughts to maintain social
acceptance. It was also recommended that blogs not be used as a means of persuasion as
grassroots influences have the most effect, and detected persuasion attempts may
undermine trust.
181
REFERENCES
Aaker, Jennifer L. (1997). “Dimensions of Brand Personality,” Journal of Marketing
Research, 34 (3), 347-356.
Abdi, Herve (2007), “Centroid, Center of Gravity, Center of Mass, Barycenter,” in
Encyclopedia of Measurement and Statistics, N.J. Salkind ed., Thousand Oaks, CA:
Sage.
Ajzen, Icek (2001), “Nature and Operation of Attitudes,” Annual Review of Psychology,
52, 27-58.
Amazon.com (2008), Statistically Improbable Phrases, (accessed April 12, 2008),
[available at http://www.amazon.com/gp/search-inside/sipshelp.html]
Anderson, Chris (2004), “The Long Tail,” Wired, 12.10, 15-20.
Arnould, Eric J., and Craig J. Thompson (2005). “Consumer Culture Theory (CCT):
Twenty Years of Research,” Journal of Consumer Research, 31 (4), 868-882.
Azouley, Audrey and Jean-Noel Kapferer (2003). “Do Brand Personality Scales Really
Measure Brand Personality,” Brand Management, 11 (2), 143-155.
Bagozzi, Richard P., Youjae Yi and Lynn W. Phillips (1991), “Assessing Construct
Validity in Organizational Research,” Administrative Science Quarterly, 36 (3), 421-
458.
Balasubramanian, Sridhar and Vijay Mahajan (2001), “The Economic Leverage of the
Virtual Community,” International Journal of Electronic Commerce, 5 (3), 103-138.
Barnett, George A. and Joseph Woelfel (1979). “On the Dimensionality of Psychological
Processes,” Quality and Quantity, 13, 215-232.
Baron, Reuben M. and David A. Kenny (1986), “The Moderator-Mediator Variable
Distinction in Social Psychological Research: Conceptual, Strategic, and Statistical
Considerations,” Journal of Personality and Social Psychology, 51 (6), 1173-1182.
Bart, Yakov, Venkatesh Shankar, Fareena Sultan and Glen L. Urban (2005), “Are the
Drivers and Role of Online Trust the Same for All Web Sites and Consumers? A
Large-Scale Exploratory Empirical Study,” Journal of Marketing, 69 (4), 133–152.
182
Belk, Russell W. (1988). “Possessions and the Extended Self,” Journal of Consumer
Research, 21 (1), 55-70.
Bellman, Richard (1961), Adaptive Control Processes: A Guided Tour, Princeton, NJ:
Princeton University Press.
Bentler, Peter M. (2005), EQS 6 Structural Equations Program Manual, Encino, CA:
Multivariate Software.
Bianchini, Monica, Marco Gori, and Franco Scarselli (2005), “Inside PageRank,” ACM
Transactions on Internet Technology, 5 (1), 92–128.
Blackmore, Susan (2001), “Evolution and Memes: The Human Brain as a Selective
Imitation Device,” Cybernetics and Systems, 32, 225-255.
Bousfield, Weston A. and Barry H. Cohen (1955). “The Occurrence of Clustering in the
Recall of Randomly Arranged Words of Different Frequencies of Usage,” Journal of
General Psychology, 30, 149-165.
----, H. William Sedgewick and Barry H. Cohen (1954), “Certain Temporal
Characteristics of the Recall of Verbal Associates,” American Journal of
Psychology, 67, 111-118.
Brown, Tom J. and Peter A. Dacin (1997). “The Company and the Product: Corporate
Associations and Consumer Product Responses,” Journal of Marketing, 61
(January), 68-84.
----, Peter A. Dacin, Michael G. Pratt and David A. Whetten (2006). “Identity, Intended
Image, Construed Image and Reputation: An Interdisciplinary Framework and
Suggested Terminology,” Journal of the Academy of Marketing Science, 34 (2), 99-
106.
Brown, Steven, Robert V. Kozinets and John F. Sherry Jr. (2003). “Teaching Old Brands
New Tricks: Retro Branding and the Revival of Brand Meaning,” Journal of
Marketing, 67 (July), 19-33.
Burnstein, Eugene and Keith Sentis (1981), “Attitude Polarization in Groups,” in
Cognitive Responses in Persuasion, Richard E. Petty, Thomas M. Ostrom and
Timothy C. Brock, eds. Hillsdale, NJ: Lawrence Erlbaum Associates, 197-216.
Burrell, Gibson and Gareth Morgan (1979), Sociological Paradigms and Organizational
Analysis, London: Heinemann.
183
Cambel, Ali Bulent (1993), Applied Chaos Theory: A Paradigm for Complexity, Boston
MA: Academic Press.
Campbell, D.T. and D.W. Fiske (1959), “Convergent and Discriminant Validation by the
Multitrait-multimethod Matrix,” Psychological Bulletin, 56, 81-105.
---- and Julian C. Stanley (1963), Experimental and Quasi-Experimental Designs for
Research, New York: Houghton Mifflin Company.
Cederman, Lars-Erik (2005), “Computational Models of Social Forms: Advancing
Generative Process Theory,” American Journal of Sociology, 110 (4), 864-893.
Chandrashekaran, Murali, Beth A. Walker, James C. Ward and Peter H. Reingen (1996),
“Modeling Individual Preference Evolution and Choice in a Dynamic Group
Setting,” Journal of Marketing Research, 33 (May), 211-223.
Chiu, Tom, DongPing Fang, John Chen, Yao Wang and Christopher Jeris (2001), “A
Robust and Scalable Clustering Algorithm for Mixed Type Attributes in Large
Database Environment,” Proceedings of the Seventh ACM SIGKDD International
Conference on Knowledge Discovery and Data Mining, 263.
Cheung, Gordon W. and Roger B. Rensvold (2002), “Evaluating Goodness-of-Fit
Indexes for Testing Measurement Invariance,” Structural Equation Modeling: A
Multidisciplinary Journal, 9 (2), 233-255.
Cialdini, Robert B. (1984), Influence: The Psychology of Persuasion, New York:
William Morrow and Company, Inc.
Cohen, Jack and Ian Stewart (1994), The Collapse of Chaos: Discovering Simplicity in a
Complex World, London: Penguin Books.
Corman, Steven and Kevin J. Dooley (2008), Crawdad Technologies, LLC, (accessed
May 11, 2008), [available at http://www.crawdadtech.com]
----, Timothy Kuhn, Robert D. McPhee and Kevin J. Dooley (2002). “Studying Complex
Discursive Systems – Centering Resonance Analysis of Communication,” Human
Communication Research, 28 (2), 157-206.
Cronbach, Lee J. (1951), “Coefficient Alpha and the Internal Structure of Tests,”
Psychometrika, 16, 297-334.
Dawkins, Richard (1976), The Selfish Gene, Oxford: Oxford University Press.
---- (2006), The God Delusion, New York: Houghton Mifflin.
184
Dellarocas, Chrysanthos (2006), “Strategic Manipulation of Internet Opinion Forums:
Implications for Consumers and Firms,” Management Science, 52 (10), 1577-1593.
Dholakia, Utpal M., Richard P. Bagozzi and Lisa Klein Pearo (2004), “A Social
Influence Model of Consumer Participation in Network- and Small-Group-Based
Virtual Communities,” International Journal of Research in Marketing, 21 (3), 241-
263.
Dowling, Grahame (2006), “How Good Corporate Reputations Create Corporate Value,”
Corporate Reputation Review, 9 (2), 134-143.
Dwyer, Paul (2007), “Measuring the Value of Electronic Word-of-Mouth and Its Impact
in Consumer Communities,” Journal of Interactive Marketing, 21 (2), 63-79.
Dwyer, Tim (2004), WilmaScope, (accessed May 11, 2008), [available at
http://wilma.sourceforge.net]
Eisenberg, Eric M. and Steven R. Phillips (1990), “What is Organizational
Miscommunication?” in John Wiemann, Nikolas Coupland and Howard Giles (eds.)
Handbook of Miscommunication and Problematic Talk, Oxford: Multilingual
Matters, 85-103.
Fornell, Claes and David F. Larcker (1981), “Evaluating Structural Equation Models
with Unobservable Variables and Measurement Error,” Journal of Marketing
Research, 18 (February), 39-50.
Fournier, Susan (1998), “Consumers and Their Brands: Developing Relationship Theory
in Consumer Research,” Journal of Consumer Research, 24 (4), 343-373.
Freling, Traci H. and Lukas P. Forbes (2005). “An Examination of Brand Personality
through Methodological Triangulation,” Brand Management, 13 (2), 148-162.
Friendly, Michael L. (1979), “In Search of the M-gram: The Structure of Organization in
Free Recall,” Cognitive Psychology, 9, 188-249.
Gintis, Herbert (2004), “Modeling Cooperation among Self-Interested Agents: A
Critique,” Journal of Socio-Economics, 33 (6), 697-717.
Gioia, Dennis A., Maijken Schultz and Kevin G. Corley (2000). “Organizational
Identity, Image, and Adaptive Instability,” Academy of Management Review, 25
(January), 63-81.
185
Godes, David, Dina Mayzlin, Yubo Chen, Sanjiv Das, Chrysanthos Dellarocas, Bruce
Pfeiffer, Barak Libai, Subrata Sen, Mengze Shi and Peeter Verlegh (2005), “The
Firm’s Management of Social Interactions,” Marketing Letters, 16 (3/4), 415-428.
Govers, Pascalle C. M. and Jan P. L. Schoormans (2005), “Product Personality and Its
Influence on Consumer Preference,” Journal of Consumer Marketing, 22 (4), 189-
197.
Granovetter, Mark (1978), “Threshold Models of Collective Behavior,” The American
Journal of Sociology, 83 (6), 1420-1443.
Grayson, Kent (1998). “The Icons of Consumer Research: Using Signs to Represent
Consumers’ Reality,” in Stern, Barbara (Ed.) Representing Consumers: Voices,
Views and Visions. London: Routledge, 27-43.
Green, Paul E., Yoram Wind and Henry J. Claycamp (1975), “Brand-Features
Congruence Mapping,” Journal of Marketing Research, 12 (August), 306-313.
Golder, Scott A. and Bernardo A. Huberman (2006), “Usage Patterns of Collaborative
Tagging Systems,” Journal of Information Science, 32 (2), 198-208.
Goldstein, Jeffery (1999), “Emergence as a Construct: History and Issues,” Emergence:
Complexity and Organization, 1, 49-72.
Hair, Joseph F., William C. Black, Barry J. Babin, Rolph E. Anderson and Ronald L.
Tatham (2005), Multivariate Data Analysis, 6th
Edition, Upper Saddle River, NJ:
Prentice Hall.
Hall, Stuart (1980). “Encoding/Decoding,” in Center for Contemporary Cultural Studies
(Ed.): Culture, Media, Language: Working Papers in Cultural Studies, 1972-79,
London: Hutchinson, 128-138.
Hardy, Melissa A. and Alan Bryman (2004), Handbook of Data Analysis, London: Sage
Publications.
Holme, Petter (2003), “Network Dynamics of Ongoing Social Relationships,”
Europhysics Letters, 64 (3), 427-433.
Howard, John A., and Jagdish N. Sheth (1969), The Theory of Buyer Behavior, New
York: John Wiley & Sons.
Hulberg, Jon (2006), “Integrating Corporate Branding and Sociological Paradigms: A
Literature Study,” Brand Management, 14 (1/2), 60-73.
186
Hutchinson, Wesley (1983), “Expertise and the Structure of Free Recall,” in Advances in
Consumer Research, Vol. 10, ed. Richard Bagozzi and Alice Tybout, Provo, UT:
Association for Consumer Research, 585-587.
Janis, Irving Lester (1972), Victims of Groupthink, Boston: Houghton Mifflin.
John, Deborah Roedder (1999), “Consumer Socialization of Children: A Retrospective
Look at Twenty-five Years of Research,” Journal of Consumer Research, 26
(December), 183-213.
Jones, Thomas O. and Earl W. Sasser Jr. (1995), “Why Satisfied Customers Defect,”
Harvard Business Review, 73 (6), 88-102.
Kamada, Tomihisa and Satoru Kawai (1989), “An Algorithm for Drawing General
Undirected Graphs,” Information Processing Letters, 31 (1), 7-15.
Kano, Noriaki (1984), “Attractive Quality and Must-Be Quality,” The Journal of the
Japanese Society for Quality Control, April, 39-48.
Kapferer, Jean-Noel (1992), Strategic Brand Management, New York: Free Press.
---- (2004), The New Strategic Brand Management: Creating and Sustaining Brand
Equity Long Term, London and Sterling, VA: Kogan Page.
Kassarjian, Harold H. (1977), “Content Analysis in Consumer Research,” Journal of
Consumer Research, 4 (June), 8-18.
Kawasaki, Guy (2004), The Art of the Start: The Time-Tested, Battle-Hardened Guide
for Anyone Starting Anything, New York: Penguin Portfolio.
Keller, Kevin L. (2003), "Brand Synthesis: The Multidimensionality of Brand
Knowledge," Journal of Consumer Research, 29 (March), 595-600.
Kerlinger, Fred H. (1964), Foundations of Behavioral Research: Education and
Psychological Inquiry, New York: Holt, Rinehart and Winston.
Kermack, William O. and Anderson G. McKendrick (1927), “A Contribution to the
Mathematical Theory of Epidemics,” Proceedings of the Royal Society of London A,
115, 700-721.
Kinkaid, D. Lawrence (1988), “The Convergence Theory and Intercultural
Communication,” in Young Y. Kim and William B. Gudykunst (eds.) Theories in
Intercultural Communication, Thousand Oaks, CA: Sage, 280-298.
187
Kirk, Roger E. (1996), “Practical Significance: A Concept Whose Time Has Come,”
Educational and Psychological Measurement, 56 (5), 746-759.
Kitchin, Robert (1998), Cyberspace the World in the Wires, Chichester: Wiley.
Kleine, Robert E., III, Susan Schultz Kleine and Jerome B. Kernan (1993), “Mundane
Consumption and the Self: A Social-Identity Perspective,” Journal of Consumer
Psychology, 2 (3), 209-235.
Kohonen, Teuvo (1990), “The Self-Organizing Map,” Proceedings of the IEEE 78 (9),
1464-1480.
Kozinets, Robert V. (2002). “The Field Behind the Screen: Using Netnography for
Marketing Research in Online Communities,” Journal of Marketing Research, 39
(February), 61-72.
Laughlin, Patrick R. and P.Christopher Earley. (1982), “Social Combination Models,
Persuasive Arguments Theory, Social Comparison Theory, and Choice Shift,”
Journal of Personality and Social Psychology, 42, 273-280.
Manzini, Ezio (1994), “Design, Environment and Social Quality: From
“Existenzminimum” to “Quality Maximum,” Design Issues, 10 (1), 37-43.
March, James G. (1991), “Exploration and Exploitation in Organizational Learning,”
Organization Science, 2, 71-87.
Mardia, Kanti V. (1970), “Measures of Multivariate Skewness and Kurtosis with
Applications,” Biometrika, 57, 519-530.
Marsden, Paul S. (1998), “Memetics: a New Paradigm for Understanding Customer
Behaviour and Influence,” Marketing Intelligence & Planning, 16 (6), 363-368.
Maslow, Abraham (1943), “A Theory of Human Motivation,” Psychological Review, 50,
370-396.
Matz, David C. and Wendy Wood (2005), “Cognitive Dissonance in Groups: The
Consequences of Disagreement,” Journal of Personality and Social Psychology, 88
(1), 22-37.
Mayzlin, Dina (2006), “Promotional Chat on the Internet,” Marketing Science, 25 (2),
155-163.
McAlexander, James H., John W Schouten, and Harold F Koening (2002), "Building
Brand Community," Journal of Marketing, 66 (January), 38-54.
188
McQueen, James B. (1967), “Some Methods for Classification and Analysis of
Multivariate Observations,” Proceedings of the Fifth Berkeley Symposium on
Mathematical Statistics and Probability, 281-297.
Merriam-Webster (2007), Merriam-Webster’s Online Dictionary, (accessed April 23,
2007), [available at http://www.m-w.com]
Mitchell, Andrew A. and Peter A. Dacin (1996), "The Assessment of Alternative
Measures of Consumer Expertise," Journal of Consumer Research, 23 (3), 219-239.
Moscovici, Serge and Marisa Zavalloni (1969), “The Group as a Polarizer of Attitudes,”
Journal of Personality and Social Psychology, 12, 125-135.
Muniz, Albert and Thomas O’Guinn (2001), “Brand Community,” Journal of Consumer
Research, 27 (March), 412-432.
Murray, Richard (2007), “The Horizontal Axis of Communication,” A Corporate Guide
to the Global Blogosphere, an Edelman White Paper (accessed May 3, 2008),
[available at
http://www.edelman.com/image/insights/content/WhitePaper011107sm.pdf]
Obloj, Tomasz and Krzysztof Obloj (2006), “Diminishing Returns from Reputation: Do
Followers Have a Competitive Advantage?” Corporate Reputation Review, 9 (4),
213-224.
Olson, Jerry C. and Aydin Muderrisoglu (1979), “The Stability of Responses Obtained
by Free Elicitation: Implications for Measuring Attribute Salience and Memory
Structure,” Advances in Consumer Research, 6 (1), 269-275.
Page, Scott (2007), The Difference: How the Power of Diversity Creates Better Groups,
Firms, Schools, and Societies, Princeton, NJ: Princeton University Press
Pereira, Paulo T., Nuno Silva and Joao Andrade e Silva (2006), “Positive and Negative
Reciprocity in the Labor Market,” Journal of Economic Behavior & Organization,
59, 406-422.
Peterson, Robert A. (2005), “Response Construction in Consumer Behavior Research,”
Journal of Business Research, 58 (3), 348-353.
Petty, Richard E. and John T. Cacioppo (1986), Communication and Persuasion:
Central and Peripheral Routes to Attitude Change, New York: Springer-Verlag.
189
----, Rao H. Unnava, and Alan J. Strathman (1991), “Theories of Attitude Change,” in
Harold H. Kassarjian and Thomas S. Robertson (Eds.), Handbook of Consumer
Behavior, New York: Prentice-Hall, 241-280.
Phelan, Steven E. (2001), “What is Complexity Science, Really?” Emergence, 3 (1),
120-136.
Pink, Daniel H. (2005), A Whole New Mind, New York: Riverhead Books.
Ratneshwar, S. (Ratti) and Allan D. Shocker, (1991), "Substitution in Use and the Role
of Usage Context in Product Category Structures," Journal of Marketing Research,
28 (August), 281-295.
Reynolds, Craig W. (1987), “Flocks, Herds, and Schools: A Distributed Behavioral
Model,” Computer Graphics, 21(4) (SIGGRAPH '87 Conference Proceedings), 25-
34.
Roberts, Peter W. and Grahame Dowling (2002), “Corporate Reputation and Sustained
Superior Financial Performance,” Strategic Management Journal, 23 (12), 1077-
1093.
Roget, Peter Mark (1962), Roget’s Thesaurus, 3th
ed. New York: Thomas Y. Crowell.
Rosa, Jose Antonio, Jelena Spanjol and Michael S. Saxon (1999), “Sociocognitive
Dynamics in a Product Market,” Journal of Marketing, 63 (Special Issue), 64-77.
Rosen, Devon (2002), “Flock Theory: Cooperative Evolution and Self-Organization of
Social Systems,” Proceedings of the 2002 CASOS (Computational Analysis of Social
and Organizational Systems) Conference. Carnegie Mellon University, Pittsburgh,
PA.
Sabate, Juan Manuel de la Fuente and Ester de Quevedo Puente (2003), “Empirical
Analysis of the Relationship between Corporate Reputation and Financial
Performance: A Survey of the Literature,” Corporate Reputation Review, 6 (2), 147-
160.
Satorra, Albert, and Peter M. Bentler (1994), “Corrections to Test Statistics and Standard
Errors in Covariance Structure Analysis,” in A. von Eye and C. C. Clogg (Eds.),
Latent Variables Analysis: Applications for Developmental Research, Thousand
Oaks, CA: Sage, 399-419.
Sawyer, R. Keith (1999), “The Emergence of Creativity,” Philosophical Psychology, 12
(4), 447-469.
190
Schwarz, Gideon, (1978), "Estimating the Dimension of a Model," Annals of Statistics, 6
(2), 461-464.
Scoble, Robert and Shel Israel (2006). Naked Conversations, Hoboken, NJ: John Wiley
& Sons
Shannon, Claude E. and Warren Weaver (1949). A Mathematical Model of
Communication. Urbana, IL: University of Illinois Press.
Shirky, Clay (2003), Power Laws, Weblogs and Inequality, (accessed May 8, 2007),
[available at http://www.shirky.com/writings/powerlaw_weblog.html]
Shoham, Aviv (2004), “Flow Experiences and Image Making: An Online Chat-Room
Ethnography,” Psychology & Marketing, 21 (10), 855-882.
Simmel, Georg and Donald Levine (1972), Georg Simmel on Individuality and Social
Forms, Chicago, IL: University of Chicago Press.
Simon, Julian (2007), Hierarchical Clustering, (accessed April 23, 2007), [available at
http://www.resample.com/xlminer/help/HClst/HClst_intro.htm]
Singer, Judith D. (1998), “Using SAS PROC MIXED to Fit Multilevel Models,
Hierarchical Models, and Individual Growth Models,” Journal of Educational and
Behavioral Statistics, 24 (4), 323-355.
Sujan, Mita (1985), "Consumer Knowledge: Effects on Evaluation Strategies Mediating
Consumer Judgment," Journal of Consumer Research, 12 (June), 31-46.
Summers, John O. (2001), “Guidelines for Conducting Research and Publishing in
Marketing: From Conceptualization Through the Review Process,” Journal of the
Academy of Marketing Science, 29 (5), 405-415.
Surowiecki, James (2004), The Wisdom of Crowds: Why the Many Are Smarter Than the
Few and How Collective Wisdom Shapes Business, Economies, Societies and
Nations, New York: Doubleday.
Technorati (2007), About Us, (accessed Aug 21, 2007), [available at
http://technorati.com/about]
Thompson, Craig J. (1997), “Interpreting Consumers: A Hermeneutical Framework for
Deriving Marketing Insights from the Texts of Consumers’ Consumption Stories,”
Journal of Marketing Research, 34 (November), 438-455.
191
Thorndike, Robert M. (1996), Measurement and Evaluation in Psychology and
Education, 6th
ed. Upper Saddle River, NJ: Prentice Hall.
Toner, John and Yuhai Tu (1998), “Flocks, Herds, and Schools: A Quantitative Theory
of Flocking,” Physical Review E, 58 (4), 4828-4858.
Van den Bulte, Christophe and Yogesh V. Joshi (2007), “New Product Diffusion with
Influentials and Imitators,” Marketing Science, 26(3), 400-421.
Vernette, Eric (2004), “Targeting Women’s Clothing Fashion Opinion Leaders in Media
Planning: An Application for Magazines,” Journal of Advertising Research, 44 (1),
90-107.
Waldrop, M. Mitchell (1992), Complexity: The Emerging Science at the Edge of Order
and Chaos, New York: Simon & Schuster.
Wallace, Patricia M. (1999), The Psychology of the Internet, Cambridge, UK:
Cambridge University Press.
Ward, James C. and Peter H. Reingen (1990), “Sociocognitive Analysis of Group
Decision Making among Consumers,” Journal of Consumer Research, 17 (3), 245-
262.
Wasserman, Stanley and Katherine Faust (1994), Social Network Analysis: Methods and
Applications, Cambridge, UK: Cambridge University Press.
Watts, Duncan J. and Peter Sheridan Dodds (2007), “Influentials, Networks, and Public
Opinion Formation,” Journal of Consumer Research, 34, 441-458.
Weick, Karl E., Kathlen M. Sutcliffe and David Obstfeld (2005), “Organizing and the
Process of Sensemaking,” Organization Science, 16 (4), 409-421.
Wikipedia (2007), Wikipedia, The Free Encyclopedia, (accessed April 23, 2007),
[available at http://en.wikipedia.org]
Woelfel, Joseph (2007), Who are we?, (accessed March 7, 2007), [available at
http://www.galileoco.com/N_who.asp]
---- and Edward L. Fink (1980), The Measurement of Communication Processes: Galileo
Theory and Method. New York: Academic Press.
Zaltman, Gerald (1982), Theory Construction in Marketing, New York: Wiley.
192
Zhang, Jin, David A. Bessler and David J. Leatham (2006), “Does Consumer Debt
Cause Economic Recession? Evidence using Directed Acyclic Graphs,” Applied
Economics Letters, 13, 401-407.
Zhang, Tian, Raghu Ramakrishnon and Miron Livny (1996). “BIRCH: An Efficient
Data Clustering Method for Very Large Databases,” Proceedings of the ACM
SIGMOD Conference on Management of Data, Montreal, Canada, 103-114.
193
APPENDIX A
GLOSSARY
Agglomerative Hierarchical Clustering: Simon (2007) describe agglomerative
hierarchical clustering as a bottom-up clustering algorithm where new clusters are
formed one-at-a-time from existing objects and clusters, generally by combining objects
or clusters closest in proximity to each other. Each step in the clustering process
becomes a level in a hierarchy. Such hierarchies are often presented in two-dimensional
diagrams known as dendrograms. An example of a dendrogram showing an
agglomerative hierarchical cluster is presented in Figure A-1.
FIGURE A-1
Dendrogram of an Agglomerative Hierarchical Cluster
194
Alignment: One of Reynolds’ (1987) three simple “steering behaviors” that
characterize flocking theory. Alignment refers to the behavior of steering toward the
average heading of local flock mates.
Allegory: One of Brown et al’s (2003) four classes of meaning that can be assigned to a
brand. Specifically, an allegory is a story with symbolic content where the brand can be
readily associated with one of the symbols and thus be endowed with meaning by the
story. For example, a financial services firm used a story called The “Tree” and I that
conveyed the message that if clients nurture the tree, a symbol for an investment, the tree
would take care of them when they got older.
Antinomy: One of Brown et al’s (2003) four classes of meaning that can be assigned to
a brand. Antinomy refers to a contradiction or paradox. The term often refers to attempts
to revive an old brand as a new product such as was done with the restyled 2005 Ford
Mustang which combined modern technology with styling reminiscent of the classic
Mustangs of the 1960’s.
Anthropomorphize: Merriam-Webster (2007) describes anthropomorphication as the
assigning of human characteristics to an inanimate object. Freling and Forbes’ (2005)
found anthropomorphication to be a common phenomenon when people form a
relationship with a brand. The brand is seen as having a personality and thus
personhood.
Arcadia: One of Brown et al’s (2003) four classes of meaning that can be assigned to a
brand. In a general sense the word “arcadia” is used to refer to a place characterized by
peace, tranquility and simplicity. It is often used to refer to someplace at a time in the
past when popular memory sees conditions as being better. A brand can be linked with
such time and place associations, such as the brand of oatmeal your grandmother served
you in the happy days of your childhood.
Associations: Merriam-Webster (2007) describes association as one or more ideas,
images, emotions or thoughts connected to something other than itself. This paper uses
the word in two ways: brand associations (thoughts connected to a brand) and cognitive
associations (the connection of different thoughts). Keller (2003) note these connections
can be the result of personal contemplation or they can be prompted by an outside
influence such as an advertisement.
Attitude: Petty et al (1991) describe attitude as a relatively enduring disposition of
thought toward a person or object, usually the result of evaluation. The word is often
used to denote a global evaluation but can be used in a more qualified sense: brand A is
good but only useful in situation X.
195
Aura: One of Brown et al’s (2003) four classes of meaning that can be assigned to a
brand. Aura refers specifically to values that a brand is thought to stand for. The term is
often applied to retro branding scenarios where old and established brands are so
associated with an adherence to old-world standards of uncompromising quality that the
brand itself conveys that meaning. It is not uncommon for an expression like “brand A is
the Rolls Royce of garbage disposals” to be used. The invocation of the Rolls Royce
automobile brand is irrelevant to garbage disposals except for the aura of almost holy
reverence the brand name evokes. Autonomy: Merriam-Webster (2007) describes autonomy as a state of being
characterized by independence and freedom. In a culture where independence is prized
over more collective concepts, it is often necessary for individuals to strike a balance
between the autonomy of unique self-expression and adherence to community standards
of acceptable expression. Flocking theory is one model of how individuals can strike
such a balance and still function effectively in helping a community achieve its goals.
Bayesian Information Criterion: An alternative name for Schwarz’ (1978) criterion,
a statistical measure used to compare alternative models proposed to explain a set of
data. The model with the lowest Schwarz criterion value fits the data best and is deemed
to be more correct.
Betweenness: An abbreviated form of the term betweenness centrality, defined
below.
Betweenness Centrality: Corman et al (2002) describes betweenness centrality as a
term from network science used to measure the degree to which a member of a network
connects otherwise disparate parts of the network together. Without this member, the
network parts (cliques or sub-networks) would be separate or substantially less
connected than they are with the member in place.
Blog: An abbreviated form of weblog, defined later.
Blogosphere: Wikipedia (2007) uses the term “blogosphere” to denote the collection
of all weblogs.
Boldness: Idea seeders and evangelists are distinguished from flockers by one general
attribute: boldness, a willingness to express thought with little regard to the opinions of
others. Boldness is used in this study as an informal construct created by combining
three parameters from equation 3.2: the number of thematic clusters started, the standard
deviation in comment-to-closest-thematic-cluster separation, and the number of
comments contributed. Boldness as an informal construct is used as an axis in Figure 3.8
to show how evangelists and idea seeders differ from flockers.
196
Brand: Kapferer (1992) notes the term “brand” is often is used to refer to concrete
symbols, such as a name, that are legally associated with an economic entity like a
company. However, these literal symbols often acquire certain cognitive associations
(thoughts, images and ideas) in the minds of people to such an extent that the thoughts,
symbols and entity are inseparable.
Brand Equity: Kapferer (1992) describes brand equity as the value associated with a
brand. It is seen as the additional money a consumer will spend to buy a branded product
over an identical unbranded product.
Brand Identity Prism: The brand identity prism is a model proposed by Kapferer
(1992) that introduced the idea of “brand personality,” a term denoting the common
human behavior of associating human personality traits with a brand.
Centering Resonance Analysis: Corman et al (2002) introduced centering
resonance analysis as a content analysis methodology where a body of text is
transformed into a network of nouns and adjectives based on the position of those words
in the sentences composing that body of text. Words can be given a measure of
betweenness centrality, called influence. Bodies of text can be compared on the basis of
their containing the same words in similar positions of influence using a measure called
resonance (each italicized word is defined in this glossary).
Centroid: Abdi (2007) describes the centroid or barycenter of a physical object as the
coordinate in n-dimensional space from which the mass of the object can be conceived
as originating. Since this proposal deals exclusively with intangible objects, the centroid
is a geometrically calculated center of a cluster of such objects.
Circuit of Communication: The circuit of communication is a model of a
communication process proposed by Hall (1980) where through an ongoing cycle of
feedback and clarification, parties in conversation converge on a shared understanding.
Cluster: Chiu et al (2001) describe a cluster as a set of objects, tangible or intangible,
that by virtue of their proximity in some dimensional space of reference are regarded as
being associable into one entity that comes to represent them all.
Cluster Feature Tree: Zhang et al (1996) describe a cluster feature tree as a compact
way of storing information about a cluster or series of nested clusters (clusters within
clusters). Its conceptualization is based on the metaphor of a tree where the objects in the
clusters are like leaves and the ways the objects are connected into clusters are like the
branches.
Cluster Sample: Wikipedia (2007) describes cluster sampling as two-stage sampling:
first you select a sample of areas of a population, and then you sample subjects within
that area. This study seeks to make general findings about the whole blogosphere. The
197
first sample is the selection of the 14 blogs in Table 4.1 from all blogs in the blogosphere
using netnography’s criterion. The second sample would be the selection of commenters
within the blogs. However, in this study all the commenters in each blog are used, except
in the pretest dataset. Cluster samples differ from random samples in that it is understood
that the subjects within each cluster are more similar than they would be in a random
sample.
Cognitive Association: Brown et al (2006) describe a cognitive association as the
establishing of a memory link between two or more objects of cognition, representing a
relationship of juxtaposition (the objects are always together) or fusion (the objects are
the same).
Cognitive Association Maps: Keller (2003) describes a cognitive association map
as a dimensional space of reference where objects of cognition can be located and their
proximity to other objects of cognition assessed.
Cognitive Diversity: In this paper cognitive diversity is used to denote the many
cognitive styles or ways individuals think, perceive and remember information and then
use that information to solve problems or, in general, enact their behavior. There are
many theories that try to explain differences in cognitive style: the nature (genetic
predisposition) versus nurture (result of formative experiences) debate is a well-known
comparison of two such theories. Contemporary organizational theory takes the
pragmatic view that regardless of why they exist, assembling teams based on cognitive
diversity results in better problem solving.
Cognitive Structure: Ward and Reingen (1990) describe a cognitive structure as an
encoding of information in the mind. This encoding can be a representation of some
tangible or intangible object, relationship, concept or body of knowledge. It can
represent one simple object, a theory of how the universe works and everything in
between. Note that a distinction is made between the brain (a physical organ) and the
mind (the result of electro-chemical reactions that form the operating system of the brain
and give us consciousness).
Cohesion: One of Reynolds’ (1987) three simple “steering behaviors” that characterize
flocking theory. Cohesion refers to the behavior of steering toward the average position
of local flock mates.
Collective Thought Diversity: Collective Thought Diversity is used as an informal
construct created by combining two parameters from equation 3.2: Mean and Stdev
Collective Thought Separation. It is designed to represent the degree of thematic
difference between individuals’ expressions of thought that are in the same collaboration
context (e.g., writing comments to the same blog entry). One way the term is used in this
study is as a latent sub-factor whose value differentiates cultural tribalism (the same
198
ideas are repeatedly echoed) from need-for-cognition (intellectual curiosity stimulates
diverse ideas).
Comments: Comments are the means whereby readers express their thoughts on the
post or entry written by the weblog author and the comments written by other readers.
Communication Convergence: A network-based model proposed by Kinkaid
(1988) that explains how people congregate about a central theme in conversation.
Kinkaid found that the density of communication networks indicates social cohesion.
Community: Kleine et al (1993) describe a community as a group of people who share
in common things (e.g., goals, beliefs, resources, preferences, needs, risks, challenges)
that reflect their individual identities. While some research may ascribe individual
identity to group membership, this study focuses on emergence, demonstrated in this
context by individuals choosing to associate based on commonalities shared prior to
group membership.
Complexity: Waldrop (1992) describes complexity as an abbreviated reference to
complex systems, mechanisms that tend to be high in dimensionality (many underlying
factors), non-linear and therefore difficult to model. The behavior of a complex system is
often attributed to emergence, a phenomenon whereby complex systems behavior is the
result of the interaction of simple systems. That is a point-of-view embraced in this
study.
Construed Image: Brown et al (2006) describe how companies try to determine how
they are viewed by consumers as a part of the process of corporate image-building
(Figure 2.1). Any such impression will be inaccurate to some extent; however, it is
desirable to minimize this inaccuracy by looking for better ways to measure it. The
image thought to be in the mind of the consumer is compared with the image the
company wants to convey on an ongoing basis. Mismatches between the two images are
used to plan future image-building activities.
Content Analysis: Kassarijian (1977) notes that content analysis is generally defined
as the study of written communications. The three common goals of content analysis are
to determine: the cognitive state (emotions, motivations, knowledge) of the writer, the
intended meaning of what was written, and the effect the writing probably had on its
readers. Content analysis usually is focused on counting the use of specific words and
their synonyms, assuming the most used words reflect the most important ideas intended
to be conveyed.
Consumer-Generated Media: Consumer-generated media is a synonym for user-
generated media, defined later.
199
Critical Thinking: Facione (2007) describes critical thinking as a form of judgment
that reflects on evidence, context and theory to understand the nature of the problem at
hand and then gathers and evaluates information needed to come to a well-justified
conclusion.
Cultural Tribalism: Kitchin (1998) describes cultural tribalism as a state that a virtual
community can be in where the members share a strong ideological identity that
welcomes the few new members who agree but eschews the many who do not. As a
result, conversation tends to be between the same people, and is thus confined to the
repetition of the same ideas; as a result, the term echo chamber is used synonymously.
Cultural tribalism evolves out of people settling into communities where the opinions
expressed match what they want to hear. Blogs characterized by cultural tribalism may
also be those where blog authors are exercising indoctrination, an ongoing attempt to
convert people to a way of thinking. What the author posts draws a homogenous group
of vocal devotees. Echo Chamber: The term “echo chamber” is a synonym for cultural tribalism, already
defined.
Eigenvalue: Measurements of complex phenomena are thought to reflect the operation
of simpler, non-divisible influences. Although these influences may be psychological in
nature, their operation is often modeled as though they were physical forces. Physical
forces are modeled as vectors, that is, they operate along a line with a magnitude and
direction. While such forces can act in any direction, when dissected into their most
simple form they are the result of a set of orthogonal (at right angles) vectors (called
eigenvectors) that act along the axes of a dimensional space. Measurements of
magnitude on theses axes are eigenvalues.
Elaboration Continuum: Petty and Cacioppo’s (1986) Elaboration Likelihood Model
of attitude change describes an elaboration continuum, a spectrum of thought depth in
response to attempts at persuasion. How deeply a target of persuasion is motivated to
think depends principally on assessments of personal relevance and need-for-cognition, a
liking of deep thinking.
Emergence: Goldstein (1999) offers a contemporary definition of emergence: “the
arising of novel and coherent structures, patterns and properties during the process of
self-organization in complex systems” (p. 51). The key factors of this definition are:
novelty (features not previously seen), coherence (self-maintaining), evolution and self-
organization (simple entities or processes collectively enact complex behaviors).
Empirical: Summers (2001) described three kinds of scientific contribution that can be
made: theoretical, methodological and empirical. The word empirical is used as an
adjective to describe results derived from real world observation or experimentation. The
word is used in contrast to “analytical”, the results of a mathematical or theoretical proof
200
and “simulated”, results derived from observing a model such as a computer program
designed to mimic a real world behavior.
Entry: Scoble and Israel (2006) use the words entry and post as synonyms for the
content a weblog author introduces to a blog to set the subject of conversation. The
author posts content in expectation that readers will reply with comments thus enabling
the author to learn from the reader community.
Entropy: Chiu et al (2001) note that “entropy” is used to refer to the loss of information
that would occur if two objects of cognition or clusters of such objects were combined
into one cluster. Stating that two objects are unique enough to persist as separate entities
implies that these objects represent dissimilar information. If two objects are judged
similar enough to combine into a cluster, and then always referenced as a cluster, some
unique information belonging to the separate objects is lost. The clustering technique
used in this study seeks to minimize such information loss.
Entropy-based Hierarchical Clustering: Entropy-based hierarchical clustering is
similar to the agglomerative hierarchical clustering already described except proximity
between objects is assessed using entropy, or information similarity, rather than some
other measure of cognitive or literal distance.
Equilibrium: Devon (2002) describes equilibrium as a state of rest due to the balance
of opposing forces. In this study, the expression of thought in groups is investigated. It is
proposed that some people will either consciously or subconsciously express their
thoughts in a manner designed to balance their desire to make a unique contribution with
their desire to support norms of community conformity.
Evangelists: In this study, evangelists are the agents in the process of mass
dissemination. They are highly prolific commenters that exhibit a low diversity of
thought expression in their comments across blog entries. They are similar to idea
seeders in that they are prolific but they differ in the narrow variation in ideas expressed.
Evolution: This study uses the term “evolution” in the same way as Dawkins (1979): a
process of change whereby a species increases its survivability from generation to
generation by slow changes to its genetic structure (the basic units of heredity) under the
influence of random mutation, procreation strategies and natural selection (survival of
the fittest). Dawkins applied the evolutionary process to explaining the way culture is
transmitted over time in his description of the meme, the idea equivalent of the gene.
Some ideas spread like a virus, mutated as they spread by people that express them in
ways that make them palatable to the hearers.
Expectation Maximization: Chiu et al (2001) describe expectation maximization as
an iterative algorithm used in a variety of applications including clustering. The
algorithm iterates through two phases: expectation and maximization. In the expectation
201
phase of clustering, the likelihood of the objects belonging to the clusters already
specified is calculated. In the maximization phase, the way objects are assigned to
clusters is altered to maximize the likelihood values calculated in the expectation step.
The two steps are iterated until no more likely way of assigning objects to clusters is
found.
Expertise: Ratneshwar and Shocker (1991) describe expertise as the possession of a
high level of knowledge and the skill needed to apply it. In this study, expertise is seen
as one type of brand association where a person has acquired a high degree of
knowledge related to the unique features of a product or product family and how to use
them. This kind of brand association can be very valuable to the holder and constitute
the basis for the way they make a living. This is a common situation for technology
products with complex functionality like database software and management information
systems.
Fixed Point Attractor: Attractors are states toward which some dynamic systems
move. A fixed point attractor is one type of attractor; it is an equilibrium state where a
system comes to rest. Cambel (1993) characterizes the movement of a system toward an
attractor to be like a firefly attracted to a light. Cohen and Stewart (1994) note that being
in an attractor state does not imply being stationary. They use as an example a table
tennis ball floating in the ocean: push the ball down it always rises to the surface (the
attractor state), however on the surface it is subject to all the forces in the environment
(e.g., wind and current).
Flocker: This study classifies commenters as idea seeders, evangelists and flockers.
Flockers are significantly more inhibited in expressing their thoughts in a group forum
than the other commenter types. They are more concerned with maintaining their
acceptance within a group and therefore control the degree to which their expressed
thought diverges from that of the group. Flocking Theory: Flocking theory is a model of how groups of birds and animals
move in groups that Rosen (2002) applied to explain how humans express their thoughts
in a group context. Every member of the flock employs three simple “steering
behaviors”: separation, alignment and cohesion which, taken together, allows each
member to be in their desired location, relative to the group, one unit of time in the
future. This model is used to explain the balance between expressing individualism and
collective solidarity that some researchers have observed in human discourse.
Forum: Wikipedia (2007) describes a forum as a website on the Internet that hosts
discussions and the posting of user-generated content. Forums often become virtual
communities as they attract a group of regular readers who enjoy contact with people
who share one or more interests. Forums are often referred to as discussion boards and
message boards.
202
Free Elicitation: “Free elicitation” is a technique for getting consumers to reveal their
thoughts introduced by researchers Olson and Muderrisoglu (1979). The technique
allows the consumer to say anything that comes to mind in response to a question
referred to as a “stimulus probe cue.”
Free Rider: Pereira et al (2006) use the term “free rider” to refer to economic actors
who consume more than their fair share of a resource while paying less than their fair
share of the cost of production. Free riding is often regarded as a problem because it can
cause essential public goods to be under-produced.
Fungible: Merriam-Webster (2007) describe the term “fungible” as an economic term
usually applied to goods that can easily be exchanged for another instance of the same
good. Fungibility is often confused with the term liquidity, the ease of exchanging a
good for money. In this study, the term is used to apply to unique associations that
consumers make with brands that make it difficult for other brands to be substituted.
Generative Process: Phelan (2001) notes that in social process theory and
complexity theory in general, phenomena that are very difficult to model and explain are
often seen as being the result of the interaction of simple processes referred to as
“generative processes.”
Generative Rules: Cederman (2005) notes that sometimes phenomena are explained
by appealing to universal or high-level laws that force certain outcomes. These
explanations are very common in explaining physical phenomena like the conversion of
friction into heat by appealing to the laws of thermodynamics. However, social
phenomena resist being robustly explained by appeals to universal laws even though
attempts to do so are common. An alternative view explains complex phenomena by
attributing them to the interaction of generative processes. When specific processes
interact, specific outcomes can be predicted as described by generative rules, well-
substantiated cause-and-effect relationships that have assumed the status of expressions
of certainty.
Grassroots: A movement driven by the spontaneous, bottom-up organizing of the
members of a community, usually around some issue. Such movements draw their
credibility from the perception that the focal issue is so important to the members of a
community that it independently motivates individuals into collective action.
Group Polarization: Burnstein and Sentis (1981) use the term “group polarization” to
refer to a migration to extreme positions during a discussion. The result is that
discussants gradually come to advocate more extreme positions than they would before
the group met. Researchers have attributed the effect to the influence of persuasive
argument combined with the desire for people to feel socially accepted. People self-
assemble into positional clusters that tend to represent extreme versions of the arguments
they find most plausible. Groupthink is often confused with group polarization and
203
indeed the two can occur together. As group polarization becomes evident within the
group, people suppress the desire to introduce information inconsistent with the position
held by the people they feel most attached to.
Groupthink: Groupthink is sometimes regarded as synonym for group polarization,
already defined. However, as Ward and Reingen (1990) note, the term describes the
specific situation where in group decision-making the members consciously strive to go
along with what they believe is a consensus. It is a situation commonly found in highly
cohesive groups where unanimity is a more important goal than properly appraising
alternatives.
Hermeneutic: Thompson (1997) describes hermeneutics as the study of text. However,
Thompson also notes it has come to be associated with the term “hermeneutic circle”
which maintains that a text cannot be interpreted without reference to its cultural, literary
(one part cannot be understood without reference to the whole) and historical contexts.
Holdout Sample: Often a dataset is divided into two parts: a training or estimation
dataset and a holdout or testing dataset. The training data set is used to estimate the
parameters of a model, while the holdout dataset is used to determine the predictive
accuracy of the model.
Hypothesis: Zaltman (1982) describes a hypothesis as a tentative assertion that
requires validation through testing. Such assertions usually propose either a causal
relationship, such as A causes B, or a softer proposition of correlation, such as A is
related to B.
Idea Seeder: “Idea seeder” is a term used in this study to denote those regular and
prolific commenters who are least constrained by any norms of conformity and therefore
freely express comments on a variety of themes, regardless of the theme of the blog
entry. These people are hypothesized to be serial starters of thematic clusters that differ
from the cluster containing the blog entry.
Identity: Brown et al (2006) use the term “identity” to denote how an organization or
person views the self. It is distinct from any use of the term “image” which denotes how
the self is viewed by others. Entities can undergo a process of image-crafting to control
how others view them. However, they also participate in identity-crafting where they
work to achieve some desired state of self as assessed by self-reflection.
Image: Gioia et al (2000) use the term “image” to refer to how, by way of cognitive
associations, others view the self. It is described as being different from the concept of
identity, defined above. There are three variations of the concept of image used in this
paper: intended, construed and actual. The actual image of some entity is only known to
the holder. The entity can construe this image and try to influence it to become more like
its intended image.
204
Individual Thought Diversity: Individual Thought Diversity is used as an informal
construct created by combining two parameters from equation 3.2: Mean and Stdev
Individual Thought Separation. It is designed to represent the freedom with which an
individual expresses thought relative to prior thoughts expressed. One way the term is
used in this study is as a latent sub-factor whose value differentiates diverse-thinking
idea seeders from the limited thinking evangelists.
FIGURE A-2
The Long Tail of the Blogosphere
Indoctrination: Dawkins (1976) note that for memes to be the cognitive equivalent of
genes and evolve under the influence of an evolutionary process, they require a means of
procreation. Dawkins (2006) proposed two basic forms of meme propagation:
indoctrination (intense repetition) and mass dissemination (spread it widely). Another
attribute of the process of indoctrination is that critical thinking is discouraged: accept
the ideology without question. Shirky (2003) observed from his survey of a wide variety
of research, that the number of readers per blog across the blogosphere has a distribution
similar to that presented in Figure A-2. A few blogs attract general interest and many
readers while most appeal to a diverse array of small readership niche interests. This
study assumes, on the basis of the same logic used when describing cultural tribalism,
205
that cognitive diversity across the blogosphere has the same distribution. As a result,
indoctrination blogs are predicted to come from the extreme end of the tail of that
distribution.
Influence: In Corman et al’s (2002) Centering Resonance Analysis (CRA), the
betweenness centrality of a word is called its influence. According to CRA, words of
high influence are those that are repeatedly used to connect other words together and are
thus the words that most reveal cognitive themes.
Inherent Classification: Chiu et al (2001) use the term “inherent classification” to
denote clustering algorithms that do not require the user to specify beforehand the
number of clusters to be found. These algorithms try to find the clusters that are
naturally present in the data.
Intangible: Arnould and Thompson (2005) note that if something is intangible it cannot
be perceived by the senses; it is incapable of being realized. They further note that many
cognitive structures are intangible.
Intended Image: Brown et al (2006) call the cognitive associations that a company
would like to instill in the minds of consumers its intended image. This image may or
may not bear a resemblance to that company’s identity, its view of itself. Both
companies and individuals often try to say that their intended image and their identity are
the same: “come and get to know the real me.”
Interclass Correlation: Muthen (1997) describes interclass correlation is a measure
of how alike one group, as a whole, is to all groups.
Interpretive Repertoire: A term introduced by Grayson (1998) to explain why
receivers may not understand the message transmitted by a sender. When a company
attempts to convey its intended image it creates a message that it can encode in a variety
of ways (Figure 2.1), unless the receiver understands the way the message is encoded
(words, figures of speech, visual metaphors, etc.) they cannot decode the message and
adopt the intended image. The resources the receiver will draw from to understand and
decode the message are termed the receiver’s interpretive repertoire.
Intraclass Correlation: Muthen (1997) describes intraclass correlation is a measure
of how alike members of a group are to each other, relative to how alike all members,
irrespective of group, are.
K-Means: McQueen (1967) introduced K-Means as a clustering algorithm that uses an
iterative approach to assigning objects to a user supplied number (K) of clusters. The
algorithm assigns the objects to K clusters at random and then calculates the centroid of
each cluster. The algorithm then reassigns objects to clusters by trying to minimize
distances between objects and cluster centroids. The algorithm is fast but not necessarily
206
good at finding the global optima. Many users repeat the algorithm several times and
then pick the best result.
Knowledge: Keller (2003) notes knowledge is one type of brand association that a
person can have. Knowledge is often related to expertise, defined above.
Latent Content Analysis: Kassarjian (1977) notes that latent and manifest content
analyses are the two main methodologies for analyzing text. Kassarjian explains that the
latent method attempts to capture the themes embedded in the text and thus is generally
considered to require human interpreters. Conclusions are based on inter-rater or inter-
coder reliability, that is, the level of agreement between the interpreters.
Long Tail: Wikipedia (2007) notes that the phrase The Long Tail was originated by
Anderson (2004) when describing the business model of firms that earn most of their
revenue catering to the needs of many people with diverse niche interests rather than
large mass markets. It is a reference to the shape of a power-law distribution, presented
in Figure A-2, where most of the phenomena depicted inhabits a long tail of low
frequency events.
Manifest Content Analysis: Kassarjian (1977) notes manifest and latent content
analyses are the two main methodologies for analyzing text. The manifest method
involves the counting of words and their frequencies, therefore ignoring the deeper
meanings of the text.
Mass Dissemination: Dawkins (2006) maintains that mass dissemination and
indoctrination are the primary memetic propagation mechanisms. This study calls the
agents of mass dissemination in the blogosphere, evangelists.
Meaning: Brown et al (2003) describe meaning as a type of cognitive association that
can be attached to a brand. There are four types of meaning that prior research has
identified: allegory (a brand story), aura (core values), arcadia (idealized community)
and antinomy (paradox: old style, new technology).
Meme: The word “meme” was introduced by Dawkins (1976) in his book The Selfish
Gene. It refers to ideas, particularly cultural norms, which are propagated from
generation to generation in a manner similar to biological genes using the evolutionary
process of random mutation, propagation and natural selection. Ideas that lead to
behaviors that help humans be more fitted to survival are continuously adapted to
circumstance and passed down through the generations as culture. The word is also
applied to ideas that people spread in the same way they might spread the flu virus: “idea
viruses.”
Memeset: Blackmore (2001) describes a memeset or memeplex as a system of related
memes that reinforce each other and increase their chances of propagation.
207
Memetics: Blackmore (2001) describes memetics is the science that studies memes and
memetic algorithms, the process of adapting memes to increase their survivability.
Blackmore notes that the process of meme propagation is different from that of
biological genes in that memes can be altered by the people that hold them. People can
alter the ideas they spread to make them more applicable to changing circumstances,
including changes in what people are willing to accept.
Message Board: Wikipedia (2007) describes a message board as a website that allows
people to hold discussions and post user-generated media. Message boards are also
called forums, as already discussed.
Methodological: Summers (2001) described three kinds of scientific contribution that
can be made: theoretical, methodological and empirical. A methodology is a process or
procedure that is accepted by a discipline as a means of knowledge discovery. A
methodological scientific contribution is the introduction of a new procedure to a
discipline. Methodological contributions must be based on accepted theory and offer
some demonstrable advantage over current methods.
Multidimensional Scaling: Green et al (1975) describe multidimensional scaling
(MDS) as a means of modeling relationships between objects that can be represented
spatially. Traditionally, MDS has only involved two- or three-dimensional
representations because they can be visualized. However, data mining technology has
allowed the detection of sophisticated patterns in data without human agency so the
dimensionality of data spaces can be unconstrained. Any object that can be expressed as
spatial coordinates using eigenvalue-based decomposition can be located and analyzed
within a multidimensional space.
Need-for-Cognition: Petty and Cacioppo (1986) described their Elaboration
Likelihood Model (ELM) of persuasion as a model of attitudinal formation and change.
When someone receives a message intended to evoke attitudinal change, message
effectiveness may well depend on triggering an intrinsic need-for-cognition motivation
in the hearer, that is, an innate enjoyment of thinking. In this study, need-for-cognition
causes people to deliberate over new ideas, expressing diverse thoughts as they try to
make sense of new ideas and fit them into their conceptual structures.
Netnography: Kozinets (2002) intoduced netnography as a methodology where the
principles of ethnography, or unobtrusive observation, were applied to the study of
virtual community. He specified a selection criterion for subject communities that
differed from the practice of standard ethnography: select communities that are focused
on a research question-relevant topic, receive above average posting traffic, have a large
number of contributing members, contain descriptively rich content and thus, in general,
enjoy a high level of member-to-member interaction.
208
Network-based Model: Wasserman and Faust (1994) note that a network-based or
simply a network model represents the relationships between objects by connecting them
in a spider web-like fashion with ties. These ties can be weighted to represent the
strength of relationships and they can be set to a length that conveys spatial proximity.
Network models can be analyzed by the methodologies that have become the tools of
network science, the study of network phenomena. A principal reference is Wasserman
and Faust’s (1994) book: Social Network Analysis.
News Group: Wikipedia (2007) describes a news group as a website that acts as a
repository for messages posted by users. News groups are similar to forums or message
boards but are regarded as being technically distinct because they are often named and
organized as a hierarchy. An example: “rec.arts.movies” where “rec” is the general
hierarchy name, “arts” and then “movies” denote a gradual narrowing of the content
subject-matter. Messages can be posted at all levels of the hierarchy based on the
author’s assessment of best fit. New levels in the hierarchy can be created by users as
needed or desired.
Noise: Shannon and Weaver (1949) describe noise as random interference that corrupts
a message while it is being transmitted between a sender and receiver. While noise exists
as a physical phenomena in any form of energy (e.g., audio, electromagnetic, heat), the
term has been used in a social and psychological sense to denote any exogenous
influence that corrupts messages in transmission. Noise can interact with endogenous
deficiencies in interpretive repertoire to cause messages to be misinterpreted.
Objects of Cognition: “Objects of cognition” is a term introduced by Woelfel and
Fink (1980) and used in this study to denote any idea, image or concept present in the
mind or stored in memory.
Organizational Associations: Brown and Dacin (1997) note that people can link
different thoughts with organizations, brands and products. It is apparent that these are
levels in a hierarchy: organizations can have many brands while many products can bear
the same branding. It is important to distinguish between the thoughts people link to the
entities at the three levels because they may be very different. For example, a person
may have a positive attitude toward one product within a brand while having a negative
attitude toward the brand or organization as a whole.
209
Outdegree: Wasserman and Faust (1994) describe outdegree as a term from network
science that refers to the number of ties an entity has initiated with other entities. For
example, if that entity is a webpage that has five (5) links to other webpages, then the
outdegree of that webpage is five (5).
Outliers. Chiu et al (2001) describe outliers as objects located far away from any
natural object clusters. Many clustering algorithms will assign outliers to the nearest
cluster but their distance to the cluster centroid is significantly greater than that of the
other objects in the cluster.
Personality: Acker (1997) describes brand personality as a type of cognitive
association that assigns human characteristics, and even identity, to a brand.
Post: Synonymous with entry, already described.
Power Law: A power-law is a class of statistical distributions where small occurrences
are common and large events are rare. One example of a power-law distribution is
presented in Figure A-2. Some examples of power-law distributions are the Pareto, Zipf
and Lognormal distributions.
Proportioning Factor: Bianchini et al (2005) note that when Google™ calculates a
web page’s PageRank™, it uses a portion of the PageRank™ of web pages that it links
to. The linking page inherits some of the importance of pages it references. This idea can
be transferred to social networks: your importance is partially based on the importance
of the people you know. Google™ keeps secret the exact proportion it uses. This paper
uses an arbitrary value of 15%.
Psychological Construct: Thorndike (1996) describes a construct as a measurable
representation of a concept. Since psychological concepts, such as trust and love, are
intangible they cannot be directly measured. Thorndike notes that to measure concepts
they must be mapped to a measurement instrument like a series of Likert scale questions.
What the measurement instrument really measures is the construct. Construct validity is
the extent to which the construct matches the concept.
Public Good: Gintis (2004) describes a public good as a resource that is not reduced
by being consumed. Blogs and forums are public goods because any number of people
can read them without diminishing the value received by any specific reader. The
problem is that public goods are not costlessly created and it is difficult to get people to
incur a cost for something that free riders can have for free. However, without some
people being willing to incur the cost, the public good cannot be created and no one gets
any value. Blogs and forums are started in the hope that as individuals benefit, enough of
them will contribute sufficient intellectual capital to ensure a positive return to everyone
regardless of the presence of free riders.
210
Quasi-experimental: Campbell and Stanley (1963) note that experiments are critical
to the empirical approach of the scientific method, aimed at supporting or falsifying
hypotheses while controlling for extraneous (unrelated to the manipulated behavior)
variation. However, some research questions require investigation in natural contexts
where extraneous variation cannot be controlled, nor can the behavior of a random
selection of subjects be manipulated while withholding such manipulation from others
(i.e., a control group). Such contexts are termed quasi-experimental. Since extraneous
variation cannot be controlled, it must be identified and its effects determined as
completely as possible.
Reciprocity: Gintis (2004) describes reciprocity as an in-kind response to the actions
of others and notes it is a standard part of most ethical systems. This study builds on
Pereira et al (2006) that found that the practice of reciprocity can be expected as long as
the price is not too high.
Recursive: Merriam-Webster (2007) describes recursion as a mathematical technique
where a function is defined in terms of itself. In this study the value of a blog entry
depends on the value of comments that are similar to it, the value of these comments also
depend on the comments similar to them, and so on.
Relationship: Fournier (1998) noted that relationship is one type of cognitive
association people often have with brands. The term implies some degree of co-
dependence and often involves the anthropomorphication (humanizing) of the brand.
Reliability: Hair et al (2005) define reliability as a “measure of the degree to which a
set of indicators of a latent construct is internally consistent in their measurements. The
indicators of highly reliable constructs are highly interrelated, indicating that they all
seem to measure the same thing.” (p. 710)
Replicator: Blackmore (2001) describes the term “replicator” within the context of
memetics as the means whereby memes are copied and dispersed. Blackmore suggests
that humanity’s main evolutionary purpose is the propagation of cultural memes.
Reputation: Gioia et al (2000) describe a brand’s reputation as the long term cognitive
associations held by consumers. Reputation is widely considered to be the most
fundamental instrument of social control. If a brand has had a good reputation for a long
time, it has probably acquired substantial brand equity as a result, the company that
owns the brand will naturally be reluctant to squander its reputation for the sake of short
term gain.
Resonance: Corman et al (2002) introduced the term “resonance” as one of the
metrics associated with Centering Resonance Analysis. Resonance is the degree of
similarity between two bodies of text based on the extent to which they both use the
same words with similar influence. Resonance therefore measures the cognitive distance
211
between two documents. Resonance can be used as a Euclidean distance in a cognitive
space.
Science of Complexity: Phelan (2001) differentiated between traditional and the
science of complexity: “Traditional science seeks direct causal relations between
elements in the universe whereas complexity theory drops down a level to explain the
rules that govern the interactions between lower-order elements that in aggregate create
emergent properties in higher-level systems” (p. 8). See the description of generative
rules discussed later.
Self-organizing Map: Kohonen (1990) introduced the self-organizing map (SOM) as
a way of visualizing a network model where the links between objects have a desired
length. If the network model is metaphorically shaken every time a new object is added
to the network, the objects will jostle around under the influence of the links that are
trying to assume their desired length until the objects assume the position relative to the
other objects that they should have. The dimensionality of the space is an important
factor in the self-organizing process. The space has to have enough dimensions to allow
the objects to move freely enough under the influence of the links to assume their proper
positions. However, it is desirable to have the minimum number of dimensions possible
to keep the space from becoming too sparse.
Sensemaking: Rosa et al (1999) found that when consumers are introduced to a new
product category they often go through a process of trying to determine what value the
category can offer them. In our interconnected world this process involves an indirect
conversation through the media and a more direct conversation through the Internet
between consumers and producers. The producers wish to clarify the nature of the
intended value offering and consumers want to assess whether one of their needs has
really been met. This process of clarification is what this study calls sensemaking.
Separation: One of Reynolds’ (1987) three simple “steering behaviors” that
characterize flocking theory. Separation refers to the behavior of steering to avoid
crowding local flock mates.
Social Attractor: The term “social attractor” is widely used in the sociology literature
in a variety of ways, although it is seldom explicitly defined. The most generic definition
seems to be Manzini’s (1994) description of something that “orients the choices of a
multiplicity of individuals” (p. 43). A social attractor is an attractor in the same sense as
a fixed point attractor. Here it is used to describe a phenomenon observed in this study
where the mean point in the spectrum of individual thought diversity draws people to an
unexpected consistency. In Figure A-3, people near the mean seem drawn together into a
commonality in the extent to which they will maintain a consistency in their thought
expression. It might be expected that the consistency with which people express thought
would be normally distributed, some few always repeating the same thoughts (i.e.,
evangelists), some few expressing a very wide diversity of thoughts (idea seeders),
212
while the majority fit neatly under the normal curve, in the middle. However, this study
observes an unexpected attraction to the mean that distorts the normal distribution. This
is discussed in greater depth in Chapter VI.
FIGURE A-3
A Social Attractor in Individual Thought Diversity
213
Social Forms: Social forms are synonymous with generative processes, already
described.
Social Process Theory: Cederman (2005) noted that the physical sciences have
many universal laws that explain observed behavior among tangible objects. Cederman
also noted that the social sciences are still looking for equivalent general theories that
explain large segments of observed phenomena. Cederman described Social Process
Theory as a promising candidate for such a general theory as it explains the complexity
of social phenomena by saying it is the emergent result of interacting generative
processes.
Thematic Cluster: This study uses self-organizing maps to locate cognitive objects in
proper proximity to each other. Once the proper locations are found, the next step is a
cluster discovery phase where it is determined whether the cognitive objects form
clusters that indicate commonality in thematic content. The number of such clusters and
their size are indications of cognitive diversity.
Theoretical: Summers (2001) described three kinds of scientific contribution that can
be made: theoretical, methodological and empirical. A theoretical contribution can be a
new explanation for relationships between variables, a new insight into the boundaries of
an existing theory or a better explanation for a relationship explained by existing theory. Threshold Model: Granovetter (1978) proposed his Threshold Model of Collective
Behavior to explain why crowds will often behave in ways the individuals in the crowd
never would on their own. He describes how for many behaviors the threshold of
individual internal resistance is normally distributed throughout a population. If the right
combination of individual thresholds is present in a crowd then behavior started by one
person can propagate through a crowd as the growing number of participants cause
thresholds of resistance to be overcome in a cascading fashion.
Trackbacks: Wikipedia (2007) notes that the entries posted by blog authors are often
referenced, through hyperlinks, in the work of other blog authors and online news
articles. These links, much like journal citations, are generally regarded as an indication
of the quality of the referenced blog entry. Such links to a blog entry are called
trackbacks.
Unconstrained Free Response: Unconstrained free response is closely connected
to the idea of free elicitation, already defined, where people are allowed to speak as
much, or as little, as the want in response to a question or “stimulus probe cue.”
User-generated Media: Wikipedia (2007) indicates that the term “user-generated
media” is used interchangeably with the terms “user-generated content” and “consumer-
generated media.” Wikipedia also indicates that the three synonym phrases refer to
online content that is produced by people who were hitherto assumed to be only users or
214
consumers of online content. The phenomenon reflects the availability of affordable new
tools for authoring content that can be easily disseminated through the Internet. Such
content includes blogs, podcasts (an audio file, usually containing commentary or
entertainment content), video, cellular phone photos, word-of-mouth and wikis (a
website that allows readers to edit the content and thus be a tool for collaborative
authoring).
Virtual Community: Wikipedia (2007) describes a virtual community as a group of
people who primarily interact via the Internet. Such communities are often very topic
oriented and therefore form around blogs and forums that the community members like
to read.
Weblog: Wikipedia (2007) describes a weblog, or blog, as a website where one or more
regular authors initiate discussion on a topic of their choosing. The website allows
comments to be added to the end of the blog author’s entry thus allowing a two-way
conversation between author and reader and a many-to-many conversation among the
readers. A weblog is usually distinguished from a forum by the existence of the blog
entry which is intended to control the subject of conversation.
Weblog Author: Scoble and Israel (2006) denote the person who writes the
conversation-starting entries for a blog as the blog author. In a corporate blog, the author
is often a high-level executive who is perceived by the readers as someone who has the
power to use the insights gained from the blog conversation.
Wikipedia: Wikipedia (2007) describes itself as a web-based encyclopedia whose
knowledge content is created and edited collectively by the users. The content of
Wikipedia is constantly in a state of flux as new information is added and existing
information is made more correct or complete by readers of varying expertise. Wikipedia
is a good example of user-generated media and an example of emergence in action.
215
APPENDIX B
LATENT VARIABLE INDICATORS
Mean Cluster Density: Mean Cluster Density is one of the overall indicators of
diversity in thought expression. After the comment clusters for each blog entry are
found, the distance from each comment’s location in cognitive space to the centroid of
its cluster can be calculated. The standard deviation of each cluster’s comment-to-
centroid distances is proposed as a measure of how dense the comments are in their
cluster. A low standard deviation indicates that the comments are closely packed and
diversity in thought expression is predicted to be low.
Mean Cluster Radius: Mean Cluster Radius is one of the overall indicators of
diversity in thought expression. After the comment clusters for each blog entry are
found, the distance from each comment’s location in cognitive space to the centroid of
its cluster can be calculated. The mean of each cluster’s comment-to-centroid distances
is proposed as a measure of how different the comments are in their cluster. A low mean
indicates that the comments are located close to their cluster centroid and diversity in
thought expression is predicted to be low.
Mean # Clusters Started: The mean number of clusters (i.e., thematic clusters)
started over the participation lifetime of a blog commenter is one of the measures used to
segment commenters into idea seeder, evangelist or flock follower categories. For each
blog entry, the commenters to that entry are grouped by category. Then, the mean
number of clusters started by each category group is one metric used to estimate the
influence of flocking, mass dissemination and idea seeding on a blog entry. Blog entries
with many commenters who have started a low number of clusters may be highly
influenced by flocking. However, blog entries whose commenters have started a large
number of clusters may be more influenced by mass dissemination or idea seeding than
flocking.
Mean Collective Thought Separation: Mean Collective Thought Separation is used
as an indicator that separates contexts characterized by cultural tribalism from those
characterized by need-for-cognition. It is the mean of entry-to-comment and comment-
to-comment cognitive distances for the current blog entry. Thus, the typical cognitive
distance between units of communication (i.e., comments and entry) is measured. It is
proposed that a low mean is indicative of little diversity in thought expression and,
moreover, a low tolerance for such diversity. This is one of the predicted attributes of a
context where cultural tribalism is taking place.
216
Mean Commenter Longevity: Mean Commenter Longevity is used as an indicator
of cultural tribalism as it is the mean time between the first comment and the current
comment of all commenters to the current blog entry. It thus measures the mean lifespan
of the active commenters to the blog.
Mean Entry-to-Entry Separation: Mean Entry-to-Entry Separation is used as an
indicator of indoctrination in blog entries and as a means of separating indoctrination
from non-indoctrination, or free thinking, blogs. It is calculated as the mean pair-wise
cognitive distances between entries to the same blog. It is predicted that if a blog author
is attempting to use indoctrination, the intense repetition of the same theme or a limited
set of themes, in a blog then the mean cognitive distance between entries will be smaller
than normal.
Mean Individual Thought Separation: Mean Individual Thought Expression is
used as an indicator whose level distinguishes evangelists from idea seeders and
contexts characterized by cultural tribalism from those characterized by need-for-
cognition. This metric is the mean comment-to-comment cognitive distance, across blog
entries, between comments made by the same individual, a commenter on the current
blog entry. Thus, the metric measures how far, in cognitive distance, an individual
commenter typically strays from a favorite theme. It is proposed that a low mean is
indicative of little willingness to express diversity in thought expression, one of the
predicted attributes of a context where cultural tribalism is taking place.
Mean # Comments: The mean number of comments (across blog entries) created by
individuals commenting on the current blog entry is proposed to be an indicator that
separates idea seeders and evangelists from the other commenters. Idea seeders and
evangelists are predicted to more prolific than flock followers.
Mean Time between Entries: Mean Time between Entries is used as an indicator of
indoctrination in blog entries and as a means of separating indoctrination from non-
indoctrination, or free thinking, blogs. It is calculated as the mean interval between
successive entries posted to a blog. It is proposed that the authors of blogs who are
employing indoctrination will post entries more often than other authors, whether
motivated by zeal or an intention to increase the intensity of the indoctrination effect.
Number of Clusters: Number of Clusters is one of the overall indicators of diversity
in thought expression. After the comment clusters for each blog entry are found, the
number of clusters is proposed to be indicative of the number of conversational themes.
It is proposed that the more clusters, the greater the diversity in thought expression.
Percentage of Outliers: Percentage of Outliers is one of the overall indicators of
diversity in thought expression. After the comment clusters for each blog entry are
found, the distance from each comment’s location in cognitive space to the centroid of
its cluster can be calculated. The mean and standard deviation of each cluster’s
217
comment-to-centroid distances is calculated and a z-score calculated for each comment.
Any comment with a z-score greater than 2.0 is considered an outlier. It is proposed that
the more outliers, the greater the diversity in thought expression.
% First Time Commenters: The number of commenters contributing to the current
blog entry who have never contributed to previous entries is predicted to have an
expansive effect on the overall diversity of thought expression in a blog. This metric is
proposed to be an indication of need-for-cognition.
% Repeat Commenters (T-1): The number of commenters contributing to the
current blog entry who contributed to the previous blog entry are predicted to impose a
limiting effect on the overall diversity of thought expression in a blog. This metric is
proposed to be an indication of cultural tribalism.
Stdev Collective Thought Separation: Stdev (i.e., standard deviation) Collective
Thought Separation is used as an indicator that separates contexts characterized by
cultural tribalism from those characterized by need-for-cognition. It is the standard
deviation of entry-to-comment and comment-to-comment cognitive distances for the
current blog entry. Thus, this metric measures the typical cognitive variation between
units of communication (i.e., comments and entry). It is proposed that a low variation is
indicative of low diversity in expressed thought, and perhaps a low tolerance for such
diversity. This is one of the predicted attributes of a context where cultural tribalism is
taking place.
Stdev Entry-to-Entry Separation: Stdev (i.e., standard deviation) Entry-to-Entry
Separation is used as an indicator of indoctrination in blog entries and as a means of
separating indoctrination from non-indoctrination, or free thinking, blogs. It is calculated
as the standard deviation in the pair-wise cognitive distances between entries to the same
blog. It is predicted that if a blog author is attempting to use indoctrination, the intense
repetition of the same theme or a limited set of themes, in a blog then the standard
deviation in cognitive distance between entries will be smaller than normal.
Stdev Individual Comment-to-Cluster Separation: The standard deviation of the
cognitive distance between a commenter’s comments and the centroids of the closest
thematic cluster across blog entries over the participation lifetime of a blog commenter is
one of the measures used to segment commenters into idea seeder, evangelist or flock
follower categories. For each blog entry, the commenters to that entry are grouped by
category. Then, the standard deviation of the comment-to-cluster separation for each
commenter in each category group is calculated as one metric used to estimate the
influence of flocking, mass dissemination and idea seeding on a blog entry. Blog entries
with commenters who have low variation in comment-to-cluster separation may be
highly influenced by flocking; as such commenters may be exercising deliberate control
over the extent to which their expressed thought differs from that of others. However,
218
blog entries whose commenters have a high variation in comment-to-cluster separation
may be more influenced by mass dissemination or idea seeding than flocking.
Stdev Individual Thought Separation: Stdev (i.e., standard deviation) Individual
Thought Expression is used as an indicator that separates contexts characterized by
cultural tribalism from those characterized by need-for-cognition. It is the standard
deviation of entry-to-comment and comment-to-comment cognitive distances for the
current blog entry. Thus, the typical variation in cognitive distances between units of
communication (i.e., comments and entry) is measured. It is proposed that a low
standard deviation is indicative of little diversity in thought expression and, moreover, a
low tolerance for such diversity. This is one of the predicted attributes of a context
where cultural tribalism is taking place. Total # Commenters: The total number of unique commenters to a blog entry is
hypothesized to determine the probability of there being an idea seeder among the
commenters.
Total Value: Total value is a metric derived from Google’s PageRank that is proposed
to measure the value of the content in the blog entry and its associated comments. The
lagged measure of this value (i.e., the total value of the previous blog entry) is proposed
to indicate reciprocation activity among commenters who feel compelled to contribute
content in return for the value they have received from the blog author and the other
commenters.
219
APPENDIX C
BLOG CORRELATION MATRICES, NORMAL PROBABILITY PLOTS AND
FULL MODELS
Blog Correlation Matrices
TABLE C-1
Correlation and Φ-matrix of Primary Constructs for AutoBlog
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.109 -
3 Flocking 0.221 0.669 -
4 IdeaSeed 0.023 0.005 0.045 -
5 Indoc -0.044 0.264 0.038 -0.022 -
6 MassDiss -0.021 0.284 0.153 0.000 0.009 -
7 NFC 0.722 0.631 0.613 0.141 -0.238 0.282 -
8 Recip 0.235 0.078 0.046 -0.024 0.149 0.005 0.035
*Note: Correlations are corrected for attenuation. All values are significant to ρ < .05.
CogDiv = Cognitive Diversity, CultTrib = Cultural Tribalism, IdeaSeed = Idea Seeding,
Indoc = Indoctrination, MassDiss = Mass Dissemination, NFC = Need-for-Cognition,
Recip = Reciprocity.
TABLE C-2
Correlation and Φ-matrix of Primary Constructs for Blog for America
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib -0.273 -
3 Flocking 0.096 0.398 -
4 IdeaSeed 0.038 -0.021 0.040 -
5 Indoc -0.087 0.242 0.012 -0.005 -
6 MassDiss 0.033 0.116 0.077 0.001 0.005 -
7 NFC 0.536 0.277 0.472 0.183 -0.245 0.187 -
8 Recip 0.261 0.052 0.026 -0.010 0.077 -0.006 0.045
220
TABLE C-3
Correlation and Φ-matrix of Primary Constructs for Blog Maverick
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib -0.180 -
3 Flocking 0.058 0.631 -
4 IdeaSeed 0.010 0.014 0.020 -
5 Indoc -0.122 0.235 -0.025 0.013 -
6 MassDiss -0.030 0.314 0.224 0.014 0.002 -
7 NFC 0.361 0.667 0.710 0.053 -0.294 0.364 -
8 Recip 0.154 0.017 0.009 0.009 0.087 0.001 -0.026
TABLE C-4
Correlation and Φ-matrix of Primary Constructs for The Consumerist
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.088 -
3 Flocking 0.294 0.437 -
4 IdeaSeed 0.073 0.008 0.100 -
5 Indoc -0.007 0.420 0.073 0.027 -
6 MassDiss 0.002 0.166 0.091 0.000 0.019 -
7 NFC 0.901 0.332 0.530 0.298 -0.069 0.129 -
8 Recip 0.221 0.055 0.031 -0.008 0.065 -0.004 0.040
221
TABLE C-5
Correlation and Φ-matrix of Primary Constructs for EnGadget
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.061 -
3 Flocking 0.260 0.551 -
4 IdeaSeed 0.057 0.023 0.051 -
5 Indoc -0.070 0.292 0.024 0.006 -
6 MassDiss -0.006 0.270 0.199 0.001 -0.011 -
7 NFC 0.738 0.715 0.677 0.153 -0.086 0.321 -
8 Recip 0.164 0.034 0.008 0.027 0.093 0.003 0.036
TABLE C-6
Correlation and Φ-matrix of Primary Constructs for The Evangelical Outpost
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.189 -
3 Flocking 0.394 0.467 -
4 IdeaSeed 0.058 0.018 0.066 -
5 Indoc 0.101 0.363 0.063 -0.014 -
6 MassDiss 0.013 0.170 0.092 -0.002 -0.016 -
7 NFC 0.995 0.492 0.626 0.232 -0.025 0.202 -
8 Recip 0.297 0.125 0.094 -0.003 0.258 -0.023 0.179
222
TABLE C-7
Correlation and Φ-matrix of Primary Constructs for GM Fastlane
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.169 -
3 Flocking 0.292 0.615 -
4 IdeaSeed 0.047 -0.004 0.063 -
5 Indoc 0.048 0.451 0.130 0.002 -
6 MassDiss -0.022 0.260 0.155 -0.011 -0.020 -
7 NFC 0.838 0.638 0.640 0.200 -0.004 0.264 -
8 Recip 0.256 0.148 0.041 -0.005 0.266 -0.011 0.137
TABLE C-8
Correlation and Φ-matrix of Primary Constructs for Freakonomics
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.204 -
3 Flocking 0.332 0.525 -
4 IdeaSeed 0.058 0.017 0.070 -
5 Indoc 0.041 0.282 0.026 -0.001 -
6 MassDiss -0.013 0.218 0.115 0.004 0.006 -
7 NFC 0.929 0.557 0.607 0.230 -0.080 0.198 -
8 Recip 0.263 0.071 0.055 -0.015 0.172 0.000 0.090
223
TABLE C-9
Correlation and Φ-matrix of Primary Constructs for Gizmodo
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib -0.204 -
3 Flocking 0.039 0.538 -
4 IdeaSeed 0.025 0.012 0.022 -
5 Indoc -0.132 0.335 0.105 0.037 -
6 MassDiss -0.018 0.221 0.126 0.000 -0.018 -
7 NFC 0.439 0.485 0.482 0.068 -0.183 0.309 -
8 Recip 0.191 0.053 0.001 0.009 0.100 -0.006 -0.010
TABLE C-10
Correlation and Φ-matrix of Primary Constructs for Google Blogoscoped
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.038 -
3 Flocking 0.437 0.397 -
4 IdeaSeed 0.113 0.005 0.124 -
5 Indoc 0.013 0.380 0.060 0.000 -
6 MassDiss 0.023 0.134 0.075 0.001 -0.001 -
7 NFC 0.977 0.367 0.704 0.318 -0.005 0.114 -
8 Recip 0.328 0.074 0.091 -0.003 0.146 0.008 0.154
224
TABLE C-11
Correlation and Φ-matrix of Primary Constructs for Joystiq
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.388 -
3 Flocking 0.460 0.537 -
4 IdeaSeed 0.042 -0.031 0.067 -
5 Indoc 0.041 0.362 0.097 0.028 -
6 MassDiss 0.034 0.283 0.184 -0.016 -0.019 -
7 NFC 0.993 0.664 0.699 0.213 -0.005 0.260 -
8 Recip 0.182 0.082 0.042 0.016 0.125 0.016 0.060
TABLE C-12
Correlation and Φ-matrix of Primary Constructs for PaulStamatiou
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib -0.033 -
3 Flocking 0.283 0.365 -
4 IdeaSeed 0.095 -0.001 0.086 -
5 Indoc 0.043 0.492 0.098 -0.020 -
6 MassDiss 0.018 0.101 0.062 -0.012 0.004 -
7 NFC 0.891 0.198 0.518 0.346 -0.031 0.073 -
8 Recip 0.231 0.079 0.044 0.008 0.217 0.023 0.084
225
TABLE C-13
Correlation and Φ-matrix of Primary Constructs for Townhall
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.056 -
3 Flocking 0.199 0.548 -
4 IdeaSeed 0.092 -0.007 0.075 -
5 Indoc -0.032 0.242 -0.006 -0.023 -
6 MassDiss -0.017 0.224 0.101 0.009 -0.013 -
7 NFC 0.727 0.590 0.576 0.269 -0.172 0.122 -
8 Recip 0.266 0.074 0.015 -0.018 0.157 0.010 0.052
TABLE C-14
Correlation and Φ-matrix of Primary Constructs for TV Squad
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.010 -
3 Flocking 0.115 0.631 -
4 IdeaSeed 0.003 -0.023 0.007 -
5 Indoc -0.073 0.244 0.036 0.010 -
6 MassDiss -0.024 0.281 0.187 -0.011 -0.030 -
7 NFC 0.519 0.624 0.517 0.024 -0.261 0.298 -
8 Recip 0.201 0.065 0.009 -0.006 0.105 0.003 0.038
226
TABLE C-15
Correlation and Φ-matrix of Primary Constructs for The Unofficial Apple Weblog
1 2 3 4 5 6 7 1 CogDiv -
2 CultTrib 0.071 -
3 Flocking 0.274 0.713 -
4 IdeaSeed 0.093 0.018 0.100 -
5 Indoc -0.113 0.341 0.086 0.007 -
6 MassDiss -0.033 0.322 0.162 -0.009 0.003 -
7 NFC 0.767 0.739 0.761 0.247 -0.102 0.200 -
8 Recip 0.221 0.088 0.041 -0.010 0.151 -0.010 0.052
FIGURE C-1
Individual Blog Normal Probability Plots of the Cognitive Diversity Factor
(Expected versus Observed Cumulative Probability)
227
FIGURE C-1 (CONT)
228
FIGURE C-2
Full Reflective and Formative Naïve Model
229
FIGURE C-3
Alternative Model of Potential Interrelationships
230
APPENDIX D
REFLECTIVE PARAMETERS FOR NAÏVE AND ALTERNATIVE MODELS
Naïve Model
FIGURE D-1
Naive Model Cognitive Diversity
FIGURE D-2
Naive Model Indoctrination / Free Thought
231
FIGURE D-3
Naive Model Cultural Tribalism
FIGURE D-4
Naive Model Idea Seeding
232
FIGURE D-5
Naive Model Flocking
FIGURE D-6
Naive Model Mass Dissemination
233
FIGURE D-7
Naive Model Reciprocation
FIGURE D-8
Naive Model Need for Cognition
234
Alternative Model
FIGURE D-9
Alternative Model Cognitive Diversity
FIGURE D-10
Alternative Model Indoctrination / Free Thought
235
FIGURE D-11
Alternative Model Cultural Tribalism
FIGURE D-12
Alternative Model Idea Seeding
236
FIGURE D-13
Alternative Model Flocking
FIGURE D-14
Alternative Model Mass Dissemination
237
FIGURE D-15
Alternative Model Need for Cognition
238
VITA
Paul Dwyer received a Bachelor of Science in Computer Engineering from the
University of South Florida in 1994, and a Master of Business Administration degree
from Texas A&M University in 2004. Prior to entering Texas A&M, he was a senior
software engineer at an Internet startup company.
His research generally focuses on consumer behavior as revealed in social
networks, word-of-mouth, participation in virtual communities, and the managerial
implications of that behavior. His work in this research stream has been published in the
Marketing Science Institute's Working Paper Series and the Journal of Interactive
Marketing, where he serves as an ad hoc referee. He has also presented his research at
American Marketing Association conferences, the International Network Science
Conference and the International Conference on Weblogs and Social Media. He was the
recipient of the Emory Marketing Institute's Doctoral Dissertation Award.
Additionally, he is interested in cognitive modeling, the emergence of complex
behavior in groups of individuals, computational modeling, data visualization and agent-
based simulation. In this vein, he was a research scholar at the Santa Fe Institute's
Complex Systems Summer School (2007).
Paul may be reached at Willamette University, Atkinson Graduate School of
Management, 900 State Street, Salem, OR 97301. His email is pdwyer@willamette.edu.
top related